{
  "_schema": 2,
  "_updated": "2026-09-11",
  "_note": "Append-only price-EVENT log rendered by /pricing/ (pricing.php via lib/pricing_data.php). Current prices are NOT here — config/pricing.json stays the single price table; the gate requires each catalogued model's last non-future event to equal its pricing.json row. Dates are the PROVIDER's effective dates (never PolyCog's catch-up date). kinds: launch | cut | increase | promo (+until) | permanent (+cancelled) | scheduled (+certainty fixed|possible) | unlisted. confidence: provider (default) | derived | secondary | approx. Refresh rule: CLAUDE.md § Standing check: the /pricing/ page. Schema 2 (Phase 64): models with status 'legacy' are not in PolyCog's catalog (retired upstream, or never/no longer offered) — they carry tier + retired and are exempt from the pricing.json gate; catalog[p].lineage {flagship, cheapest} + api_launch drive the since-launch charts. Legacy entries without a retired date carry last_checked (the day the provider still listed them); the gate warns when that is over 12 months old.",
  "_verified": {
    "openai": "2026-09-11",
    "anthropic": "2026-09-11",
    "gemini": "2026-09-11",
    "deepseek": "2026-09-11",
    "xai": "2026-09-11"
  },
  "catalog": {
    "openai": {
      "name": "ChatGPT",
      "vendor": "OpenAI",
      "source": "https://developers.openai.com/api/docs/pricing",
      "source_name": "OpenAI pricing page",
      "featured": 4,
      "default_top": "gpt-6-astra",
      "default_initial": "gpt-5.6-luna",
      "sample_shared": [],
      "synthesis_eligible": true,
      "models": [
        "gpt-6-astra",
        "gpt-5.6-sol",
        "gpt-5.6-terra",
        "gpt-5.6-luna",
        "gpt-5.5",
        "gpt-5.4",
        "gpt-5.1",
        "gpt-4.1",
        "gpt-5.4-mini",
        "gpt-4.1-mini",
        "gpt-4o-mini"
      ],
      "api_launch": {
        "date": "2020-06-11",
        "note": "OpenAI API opened in private beta (subscription/access-tier pricing, not yet a public per-token price list). The first public per-token price list for the GPT-3 engines (davinci/curie/babbage/ada) went live around September 2021, and the waitlist was removed entirely on 2021-11-18.",
        "source": "https://techcrunch.com/2020/06/11/openai-makes-an-all-purpose-api-for-its-text-based-ai-capabilities/",
        "confidence": "secondary"
      },
      "lineage": {
        "flagship": [
          {
            "id": "davinci",
            "from": "2021-09-01"
          },
          {
            "id": "text-davinci-003",
            "from": "2022-11-28"
          },
          {
            "id": "gpt-4",
            "from": "2023-03-14"
          },
          {
            "id": "gpt-4-turbo",
            "from": "2023-11-06"
          },
          {
            "id": "gpt-4o",
            "from": "2024-05-13"
          },
          {
            "id": "o1",
            "from": "2024-12-17"
          },
          {
            "id": "o3",
            "from": "2025-04-16"
          },
          {
            "id": "gpt-5",
            "from": "2025-08-07"
          },
          {
            "id": "gpt-5.1",
            "from": "2025-11-13"
          },
          {
            "id": "gpt-5.2",
            "from": "2025-12-11"
          },
          {
            "id": "gpt-5.4",
            "from": "2026-03-05"
          },
          {
            "id": "gpt-5.5",
            "from": "2026-04-24"
          },
          {
            "id": "gpt-5.6-sol",
            "from": "2026-07-09"
          },
          {
            "id": "gpt-6-astra",
            "from": "2026-09-03"
          }
        ],
        "cheapest": [
          {
            "id": "ada",
            "from": "2021-09-01"
          },
          {
            "id": "gpt-3.5-turbo",
            "from": "2024-01-04"
          },
          {
            "id": "gpt-4o-mini",
            "from": "2024-07-18"
          },
          {
            "id": "gpt-4.1-nano",
            "from": "2025-04-14"
          },
          {
            "id": "gpt-5-nano",
            "from": "2025-08-07"
          }
        ]
      }
    },
    "anthropic": {
      "name": "Claude",
      "vendor": "Anthropic",
      "source": "https://platform.claude.com/docs/en/about-claude/pricing",
      "source_name": "Claude pricing page",
      "featured": 4,
      "default_top": "claude-fable-5-1",
      "default_initial": "claude-sonnet-5",
      "sample_shared": [
        "claude-haiku-4-5-20251001"
      ],
      "synthesis_eligible": true,
      "models": [
        "claude-fable-5-1",
        "claude-opus-5",
        "claude-sonnet-5",
        "claude-haiku-4-5-20251001",
        "claude-fable-5",
        "claude-opus-4-8",
        "claude-opus-4-7",
        "claude-opus-4-6",
        "claude-sonnet-4-6"
      ],
      "api_launch": {
        "date": "2023-03-14",
        "note": "\"Introducing Claude\": Claude and Claude Instant offered via API in the developer console, access on request (waitlist). Context window was 9K tokens until 2023-05-11 (expanded to 100K). The API was declared generally available (no waitlist) with the Claude 3 launch on 2024-03-04 (\"our API, which is now generally available\").",
        "source": "https://www.anthropic.com/news/introducing-claude",
        "confidence": "provider"
      },
      "lineage": {
        "flagship": [
          {
            "id": "claude-1",
            "from": "2023-03-14"
          },
          {
            "id": "claude-2.0",
            "from": "2023-07-11"
          },
          {
            "id": "claude-2.1",
            "from": "2023-11-21"
          },
          {
            "id": "claude-3-opus-20240229",
            "from": "2024-03-04"
          },
          {
            "id": "claude-3-5-sonnet-20240620",
            "from": "2024-06-20"
          },
          {
            "id": "claude-3-5-sonnet-20241022",
            "from": "2024-10-22"
          },
          {
            "id": "claude-3-7-sonnet-20250219",
            "from": "2025-02-24"
          },
          {
            "id": "claude-opus-4-20250514",
            "from": "2025-05-22"
          },
          {
            "id": "claude-opus-4-1-20250805",
            "from": "2025-08-05"
          },
          {
            "id": "claude-sonnet-4-5-20250929",
            "from": "2025-09-29"
          },
          {
            "id": "claude-opus-4-5-20251101",
            "from": "2025-11-24"
          },
          {
            "id": "claude-opus-4-6",
            "from": "2026-02-05"
          },
          {
            "id": "claude-opus-4-7",
            "from": "2026-04-16"
          },
          {
            "id": "claude-opus-4-8",
            "from": "2026-05-28"
          },
          {
            "id": "claude-fable-5",
            "from": "2026-06-09"
          },
          {
            "id": "claude-fable-5-1",
            "from": "2026-09-01"
          }
        ],
        "flagship_alt_top_tier": [
          {
            "id": "claude-1",
            "from": "2023-03-14"
          },
          {
            "id": "claude-2.0",
            "from": "2023-07-11"
          },
          {
            "id": "claude-2.1",
            "from": "2023-11-21"
          },
          {
            "id": "claude-3-opus-20240229",
            "from": "2024-03-04"
          },
          {
            "id": "claude-opus-4-20250514",
            "from": "2025-05-22"
          },
          {
            "id": "claude-opus-4-1-20250805",
            "from": "2025-08-05"
          },
          {
            "id": "claude-opus-4-5-20251101",
            "from": "2025-11-24"
          },
          {
            "id": "claude-opus-4-6",
            "from": "2026-02-05"
          },
          {
            "id": "claude-opus-4-7",
            "from": "2026-04-16"
          },
          {
            "id": "claude-opus-4-8",
            "from": "2026-05-28"
          },
          {
            "id": "claude-fable-5",
            "from": "2026-06-09"
          },
          {
            "id": "claude-fable-5-1",
            "from": "2026-09-01"
          }
        ],
        "cheapest": [
          {
            "id": "claude-instant-1",
            "from": "2023-03-14"
          },
          {
            "id": "claude-3-haiku-20240307",
            "from": "2024-03-13"
          },
          {
            "id": "claude-haiku-4-5-20251001",
            "from": "2026-04-20"
          }
        ],
        "notes": "`flagship` follows Anthropic's own 'our most intelligent / best model' claim at each launch, which twice put a $3/$15 Sonnet above the then-current $15/$75 Opus (3.5 Sonnet on 2024-06-20 over Claude 3 Opus; Sonnet 4.5 on 2025-09-29 over Opus 4.1). `flagship_alt_top_tier` is the alternative reading (top price tier only: Claude 1/2/2.1 -> Opus line -> Fable line) if the chart should not dip during those windows. Opus 5 (2026-07-24, $5/$25) is omitted from both because Fable 5 (2026-06-09, $10/$50) remained 'our most capable widely released model'. `cheapest`: Claude Instant 1.x until Claude 3 Haiku shipped; Claude 3 Haiku ($0.25/$1.25) stayed the cheapest GA model — Claude 3.5 Haiku ($1/$5, then $0.80/$4) and Haiku 4.5 ($1/$5) were never cheaper — until it was retired on 2026-04-20, after which Haiku 4.5 is cheapest."
      }
    },
    "gemini": {
      "name": "Gemini",
      "vendor": "Google",
      "source": "https://ai.google.dev/gemini-api/docs/pricing",
      "source_name": "Gemini API pricing page",
      "featured": 5,
      "default_top": "gemini-3.1-pro-preview",
      "default_initial": "gemini-3.8-flash",
      "sample_shared": [
        "gemini-2.5-flash-lite"
      ],
      "synthesis_eligible": true,
      "models": [
        "gemini-3.1-pro-preview",
        "gemini-3.8-flash",
        "gemini-3.7-flash",
        "gemini-3-flash-preview",
        "gemini-2.5-flash-lite",
        "gemini-3.6-flash",
        "gemini-3.5-flash",
        "gemini-3.5-flash-lite",
        "gemini-3.1-flash-lite"
      ],
      "api_launch": {
        "date": "2023-12-13",
        "note": "\"It's time for developers and enterprises to build with Gemini Pro\": Gemini Pro (text) and Gemini Pro Vision (multimodal) released in public preview via Google AI Studio and Vertex AI, free during preview (\"free to use right now, within limits\", 60 RPM). The post says paid pricing (\"per 1,000 characters or per image\") would follow \"early next year\". Gemini 1.0 Pro reached general availability with",
        "source": "https://developers.googleblog.com/2023/12/build-with-gemini-pro.html",
        "confidence": "provider"
      },
      "lineage": {
        "flagship": [
          {
            "id": "gemini-1.0-pro",
            "from": "2024-02-15"
          },
          {
            "id": "gemini-1.5-pro",
            "from": "2024-05-23"
          },
          {
            "id": "gemini-2.5-pro",
            "from": "2025-04-04"
          },
          {
            "id": "gemini-3-pro-preview",
            "from": "2025-11-18"
          },
          {
            "id": "gemini-3.1-pro-preview",
            "from": "2026-02-19"
          }
        ],
        "cheapest": [
          {
            "id": "gemini-1.0-pro",
            "from": "2024-02-15"
          },
          {
            "id": "gemini-1.5-flash",
            "from": "2024-05-23"
          },
          {
            "id": "gemini-1.5-flash-8b",
            "from": "2024-10-03"
          },
          {
            "id": "gemini-2.0-flash-lite",
            "from": "2025-09-29"
          },
          {
            "id": "gemini-2.5-flash-lite",
            "from": "2026-06-01"
          }
        ]
      }
    },
    "deepseek": {
      "name": "DeepSeek",
      "vendor": "DeepSeek",
      "source": "https://api-docs.deepseek.com/quick_start/pricing",
      "source_name": "DeepSeek pricing page",
      "featured": 2,
      "default_top": "deepseek-v4-pro",
      "default_initial": "deepseek-flash",
      "sample_shared": [
        "deepseek-flash"
      ],
      "synthesis_eligible": true,
      "models": [
        "deepseek-v4-pro",
        "deepseek-flash"
      ],
      "api_launch": {
        "date": "2024-05-06",
        "note": "DeepSeek-V2 (deepseek-chat) API launch — DeepSeek's first publicly priced API tier: ¥1 per 1M input tokens / ¥2 per 1M output tokens (≈$0.14/$0.28), pitched publicly as roughly 100x cheaper than GPT-4-Turbo. No context-caching tier existed yet (added Aug 2024).",
        "source": "https://github.com/deepseek-ai/DeepSeek-V2",
        "confidence": "secondary"
      },
      "lineage": {
        "flagship": [
          {
            "id": "deepseek-chat",
            "from": "2024-05-06"
          },
          {
            "id": "deepseek-reasoner",
            "from": "2025-01-20"
          },
          {
            "id": "deepseek-v4-pro",
            "from": "2026-04-24"
          }
        ],
        "cheapest": [
          {
            "id": "deepseek-chat",
            "from": "2024-05-06"
          },
          {
            "id": "deepseek-v4-flash",
            "from": "2026-04-24"
          },
          {
            "id": "deepseek-flash",
            "from": "2026-09-10"
          }
        ]
      }
    },
    "xai": {
      "name": "Grok",
      "vendor": "xAI",
      "source": "https://docs.x.ai/developers/pricing",
      "source_name": "xAI pricing page",
      "featured": 3,
      "default_top": "grok-4.6",
      "default_initial": "grok-4.3",
      "sample_shared": [],
      "synthesis_eligible": false,
      "models": [
        "grok-4.6",
        "grok-4.5",
        "grok-4.3"
      ],
      "api_launch": {
        "date": "2024-10-21",
        "note": "xAI opened paid API access with a single text model, grok-beta (128K context, function calling, system prompts), at console.x.ai / api.x.ai, OpenAI- and Anthropic-SDK compatible. xAI's own blog post formalizing this as a 'Public Beta' and detailing $25/month free credits is dated November 4, 2024 (2 weeks later) — both dates are cited below; the earlier date is used as the API launch per contempor",
        "source": "https://techcrunch.com/2024/10/21/xai-elon-musks-ai-startup-launches-an-api",
        "confidence": "secondary"
      },
      "lineage": {
        "flagship": [
          {
            "id": "grok-beta",
            "from": "2024-10-21"
          },
          {
            "id": "grok-2-1212",
            "from": "2024-12-14"
          },
          {
            "id": "grok-3",
            "from": "2025-04-09"
          },
          {
            "id": "grok-4",
            "from": "2025-07-09"
          },
          {
            "id": "grok-4-1",
            "from": "2025-11-17"
          },
          {
            "id": "grok-4.20-0309-reasoning",
            "from": "2026-03-10"
          },
          {
            "id": "grok-4.3",
            "from": "2026-05-01"
          },
          {
            "id": "grok-4.5",
            "from": "2026-07-16"
          },
          {
            "id": "grok-4.6",
            "from": "2026-08-12"
          }
        ],
        "cheapest": [
          {
            "id": "grok-beta",
            "from": "2024-10-21"
          },
          {
            "id": "grok-2-1212",
            "from": "2024-12-14"
          },
          {
            "id": "grok-3-mini",
            "from": "2025-04-09"
          },
          {
            "id": "grok-4-fast-non-reasoning",
            "from": "2025-09-19"
          },
          {
            "id": "grok-4.3",
            "from": "2026-05-15"
          }
        ]
      }
    }
  },
  "models": {
    "gpt-6-astra": {
      "name": "GPT-6 Astra",
      "provider": "openai",
      "events": [
        {
          "date": "2026-09-03",
          "in": 10,
          "out": 50,
          "kind": "launch",
          "note": "Cached input $1.00. Prompts over 272K tokens bill 2× input / 1.5× output.",
          "source": "https://developers.openai.com/api/docs/changelog"
        }
      ],
      "tier": "flagship"
    },
    "gpt-5.6-sol": {
      "name": "GPT-5.6 Sol",
      "provider": "openai",
      "events": [
        {
          "date": "2026-07-09",
          "in": 5,
          "out": 30,
          "kind": "launch",
          "source": "https://developers.openai.com/api/docs/changelog"
        },
        {
          "date": "2026-08-21",
          "in": 4,
          "out": 20,
          "kind": "promo",
          "until": "2026-11-21",
          "note": "OpenAI: “promotional pricing is available at least through November 21, 2026.”",
          "source": "https://developers.openai.com/api/docs/changelog"
        },
        {
          "date": "2026-11-22",
          "in": 5,
          "out": 30,
          "kind": "scheduled",
          "certainty": "possible",
          "note": "Only if the promo lapses — OpenAI has not announced an end price; the pre-promo list was $5 / $30."
        }
      ],
      "polycog_notes": [
        {
          "date": "2026-09-02",
          "approx": false,
          "text": "PolyCog’s estimate stayed at $5 / $30 from Aug 21 until Sep 2, 2026 (over-estimated 20–33%)."
        }
      ],
      "tier": "flagship"
    },
    "gpt-5.6-terra": {
      "name": "GPT-5.6 Terra",
      "provider": "openai",
      "events": [
        {
          "date": "2026-07-09",
          "in": 2.5,
          "out": 15,
          "kind": "launch",
          "source": "https://developers.openai.com/api/docs/changelog"
        },
        {
          "date": "2026-07-30",
          "in": 2,
          "out": 12,
          "kind": "cut",
          "note": "OpenAI: “Starting July 30, GPT-5.6 Luna costs 80% less, while GPT-5.6 Terra costs 20% less.”",
          "source": "https://developers.openai.com/api/docs/changelog"
        }
      ],
      "polycog_notes": [
        {
          "date": "2026-09-02",
          "approx": false,
          "text": "PolyCog’s estimate stayed at $2.50 / $15 from Jul 30 until Sep 2, 2026 (over-estimated 25%)."
        }
      ],
      "tier": "mid"
    },
    "gpt-5.6-luna": {
      "name": "GPT-5.6 Luna",
      "provider": "openai",
      "events": [
        {
          "date": "2026-07-09",
          "in": 1,
          "out": 6,
          "kind": "launch",
          "source": "https://developers.openai.com/api/docs/changelog"
        },
        {
          "date": "2026-07-30",
          "in": 0.2,
          "out": 1.2,
          "kind": "cut",
          "note": "An 80% permanent cut, announced with Terra’s.",
          "source": "https://developers.openai.com/api/docs/changelog"
        }
      ],
      "polycog_notes": [
        {
          "date": "2026-09-02",
          "approx": false,
          "text": "PolyCog’s estimate stayed at $1 / $6 from Jul 30 until Sep 2, 2026 (over-estimated 5×)."
        }
      ],
      "tier": "small"
    },
    "gpt-5.5": {
      "name": "GPT-5.5",
      "provider": "openai",
      "events": [
        {
          "date": "2026-04-24",
          "in": 5,
          "out": 30,
          "kind": "launch",
          "source": "https://developers.openai.com/api/docs/changelog"
        }
      ],
      "tier": "flagship"
    },
    "gpt-5.4": {
      "name": "GPT-5.4",
      "provider": "openai",
      "events": [
        {
          "date": "2026-03-05",
          "in": 2.5,
          "out": 15,
          "kind": "launch",
          "source": "https://developers.openai.com/api/docs/changelog"
        }
      ],
      "tier": "flagship"
    },
    "gpt-5.1": {
      "name": "GPT-5.1",
      "provider": "openai",
      "events": [
        {
          "date": "2025-11-13",
          "in": 1.25,
          "out": 10,
          "kind": "launch",
          "source": "https://developers.openai.com/api/docs/changelog"
        }
      ],
      "tier": "flagship"
    },
    "gpt-4.1": {
      "name": "GPT-4.1",
      "provider": "openai",
      "events": [
        {
          "date": "2025-04-14",
          "in": 2,
          "out": 8,
          "kind": "launch",
          "source": "https://developers.openai.com/api/docs/changelog"
        }
      ],
      "tier": "mid"
    },
    "gpt-5.4-mini": {
      "name": "GPT-5.4 mini",
      "provider": "openai",
      "events": [
        {
          "date": "2026-03-17",
          "in": 0.75,
          "out": 4.5,
          "kind": "launch",
          "source": "https://developers.openai.com/api/docs/changelog"
        }
      ],
      "tier": "mid"
    },
    "gpt-4.1-mini": {
      "name": "GPT-4.1 mini",
      "provider": "openai",
      "events": [
        {
          "date": "2025-04-14",
          "in": 0.4,
          "out": 1.6,
          "kind": "launch",
          "source": "https://developers.openai.com/api/docs/changelog"
        }
      ],
      "tier": "small"
    },
    "gpt-4o-mini": {
      "name": "GPT-4o mini",
      "provider": "openai",
      "events": [
        {
          "date": "2024-07-18",
          "in": 0.15,
          "out": 0.6,
          "kind": "launch",
          "source": "https://developers.openai.com/api/docs/models/gpt-4o-mini"
        }
      ],
      "tier": "small"
    },
    "claude-fable-5-1": {
      "name": "Claude Fable 5.1",
      "provider": "anthropic",
      "events": [
        {
          "date": "2026-09-01",
          "in": 10,
          "out": 50,
          "kind": "launch",
          "note": "Same rate card as Fable 5; cache reads cut to $0.25 per MTok.",
          "source": "https://platform.claude.com/docs/en/release-notes/overview"
        }
      ],
      "tier": "flagship"
    },
    "claude-opus-5": {
      "name": "Claude Opus 5",
      "provider": "anthropic",
      "events": [
        {
          "date": "2026-07-24",
          "in": 5,
          "out": 25,
          "kind": "launch",
          "source": "https://platform.claude.com/docs/en/release-notes/overview"
        }
      ],
      "tier": "flagship"
    },
    "claude-sonnet-5": {
      "name": "Claude Sonnet 5",
      "provider": "anthropic",
      "events": [
        {
          "date": "2026-06-30",
          "in": 2,
          "out": 10,
          "kind": "launch",
          "promo": true,
          "until": "2026-08-31",
          "note": "Launched at an introductory $2 / $10, scheduled to rise to $3 / $15 on Sep 1, 2026.",
          "source": "https://platform.claude.com/docs/en/release-notes/overview"
        },
        {
          "date": "2026-08-10",
          "in": 2,
          "out": 10,
          "kind": "permanent",
          "cancelled": {
            "in": 3,
            "out": 15
          },
          "note": "Anthropic made the introductory price the standard price; the scheduled Sep 1 increase to $3 / $15 will not occur.",
          "source": "https://platform.claude.com/docs/en/release-notes/overview"
        }
      ],
      "polycog_notes": [
        {
          "date": "2026-09-02",
          "approx": false,
          "text": "PolyCog estimated at the announced $3 / $15 list price (a deliberate 1.5× over-estimate) from mid-July until Sep 2, 2026."
        }
      ],
      "tier": "mid"
    },
    "claude-haiku-4-5-20251001": {
      "name": "Claude Haiku 4.5",
      "provider": "anthropic",
      "events": [
        {
          "date": "2025-10-15",
          "in": 1,
          "out": 5,
          "kind": "launch",
          "confidence": "derived",
          "note": "Eligible for retirement not sooner than Oct 15, 2026 (no notice issued).",
          "source": "https://platform.claude.com/docs/en/about-claude/model-deprecations"
        }
      ],
      "polycog_notes": [
        {
          "date": "2026-07-21",
          "approx": true,
          "text": "PolyCog’s table carried $0.80 / $4.00 (a mis-entry) until mid-July 2026 — estimates were 20% low."
        }
      ],
      "tier": "small"
    },
    "claude-fable-5": {
      "name": "Claude Fable 5",
      "provider": "anthropic",
      "events": [
        {
          "date": "2026-06-09",
          "in": 10,
          "out": 50,
          "kind": "launch",
          "source": "https://platform.claude.com/docs/en/release-notes/overview"
        }
      ],
      "tier": "flagship"
    },
    "claude-opus-4-8": {
      "name": "Claude Opus 4.8",
      "provider": "anthropic",
      "events": [
        {
          "date": "2026-05-28",
          "in": 5,
          "out": 25,
          "kind": "launch",
          "source": "https://platform.claude.com/docs/en/release-notes/overview"
        }
      ],
      "tier": "flagship"
    },
    "claude-opus-4-7": {
      "name": "Claude Opus 4.7",
      "provider": "anthropic",
      "events": [
        {
          "date": "2026-04-16",
          "in": 5,
          "out": 25,
          "kind": "launch",
          "confidence": "derived",
          "source": "https://platform.claude.com/docs/en/about-claude/model-deprecations"
        }
      ],
      "tier": "flagship"
    },
    "claude-opus-4-6": {
      "name": "Claude Opus 4.6",
      "provider": "anthropic",
      "events": [
        {
          "date": "2026-02-05",
          "in": 5,
          "out": 25,
          "kind": "launch",
          "confidence": "derived",
          "source": "https://platform.claude.com/docs/en/about-claude/model-deprecations"
        }
      ],
      "polycog_notes": [
        {
          "date": "2026-07-21",
          "approx": true,
          "text": "PolyCog’s table carried $15 / $75 (a mis-entry) until mid-July 2026 — restored Opus 4.6 sessions were over-reported 3×."
        }
      ],
      "tier": "flagship"
    },
    "claude-sonnet-4-6": {
      "name": "Claude Sonnet 4.6",
      "provider": "anthropic",
      "events": [
        {
          "date": "2026-02-17",
          "in": 3,
          "out": 15,
          "kind": "launch",
          "confidence": "derived",
          "source": "https://platform.claude.com/docs/en/about-claude/model-deprecations"
        }
      ],
      "tier": "mid"
    },
    "gemini-3.1-pro-preview": {
      "name": "Gemini 3.1 Pro (preview)",
      "provider": "gemini",
      "events": [
        {
          "date": "2026-02-19",
          "in": 2,
          "out": 12,
          "kind": "launch",
          "note": "Prompts over 200K tokens bill $4 / $18.",
          "source": "https://ai.google.dev/gemini-api/docs/changelog"
        }
      ],
      "tier": "flagship"
    },
    "gemini-3.8-flash": {
      "name": "Gemini 3.8 Flash",
      "provider": "gemini",
      "events": [
        {
          "date": "2026-09-02",
          "in": 0.75,
          "out": 3.75,
          "kind": "launch",
          "promo": true,
          "until": "2026-12-31",
          "note": "Launched at Google’s introductory Flash rate, “through December 31, 2026.”",
          "source": "https://ai.google.dev/gemini-api/docs/changelog"
        },
        {
          "date": "2027-01-01",
          "in": 1.5,
          "out": 7.5,
          "kind": "scheduled",
          "certainty": "fixed",
          "note": "Google: “$1.50 / $7.50 starting January 1, 2027.”",
          "source": "https://ai.google.dev/gemini-api/docs/pricing"
        }
      ],
      "tier": "mid"
    },
    "gemini-3.7-flash": {
      "name": "Gemini 3.7 Flash",
      "provider": "gemini",
      "events": [
        {
          "date": "2026-08-13",
          "in": 0.75,
          "out": 3.75,
          "kind": "launch",
          "promo": true,
          "until": "2026-12-31",
          "note": "Launched “at an introductory price through December 31, 2026.”",
          "source": "https://ai.google.dev/gemini-api/docs/changelog"
        },
        {
          "date": "2027-01-01",
          "in": 1.5,
          "out": 7.5,
          "kind": "scheduled",
          "certainty": "fixed",
          "source": "https://ai.google.dev/gemini-api/docs/pricing"
        }
      ],
      "tier": "mid"
    },
    "gemini-3-flash-preview": {
      "name": "Gemini 3 Flash (preview)",
      "provider": "gemini",
      "events": [
        {
          "date": "2025-12-17",
          "in": 0.5,
          "out": 3,
          "kind": "launch",
          "note": "Google now calls it its legacy Flash model.",
          "source": "https://techcrunch.com/2025/12/17/google-launches-gemini-3-flash-makes-it-the-default-model-in-the-gemini-app/"
        }
      ],
      "tier": "mid"
    },
    "gemini-2.5-flash-lite": {
      "name": "Gemini 2.5 Flash-Lite",
      "provider": "gemini",
      "events": [
        {
          "date": "2025-07-22",
          "in": 0.1,
          "out": 0.4,
          "kind": "launch",
          "source": "https://ai.google.dev/gemini-api/docs/changelog"
        },
        {
          "date": "2026-09-05",
          "in": 0.1,
          "out": 0.4,
          "kind": "unlisted",
          "note": "Google removed the model from its pricing page and, from Sep 6, closed it to new Google Cloud projects; existing keys are still served at the last listed price.",
          "source": "https://ai.google.dev/gemini-api/docs/pricing"
        }
      ],
      "tier": "small"
    },
    "gemini-3.6-flash": {
      "name": "Gemini 3.6 Flash",
      "provider": "gemini",
      "events": [
        {
          "date": "2026-07-21",
          "in": 1.5,
          "out": 7.5,
          "kind": "launch",
          "source": "https://ai.google.dev/gemini-api/docs/changelog"
        },
        {
          "date": "2026-08-13",
          "in": 0.75,
          "out": 3.75,
          "kind": "promo",
          "until": "2026-12-31",
          "confidence": "secondary",
          "note": "Moved onto the introductory Flash rate the day 3.7 Flash launched. No changelog entry; dated from pricing-page captures.",
          "source": "https://ai.google.dev/gemini-api/docs/pricing"
        },
        {
          "date": "2027-01-01",
          "in": 1.5,
          "out": 7.5,
          "kind": "scheduled",
          "certainty": "fixed",
          "source": "https://ai.google.dev/gemini-api/docs/pricing"
        }
      ],
      "polycog_notes": [
        {
          "date": "2026-09-02",
          "approx": false,
          "text": "PolyCog’s estimate stayed at $1.50 / $7.50 from Aug 13 until Sep 2, 2026 (over-estimated 2×)."
        }
      ],
      "tier": "mid"
    },
    "gemini-3.5-flash": {
      "name": "Gemini 3.5 Flash",
      "provider": "gemini",
      "events": [
        {
          "date": "2026-05-19",
          "in": 1.5,
          "out": 9,
          "kind": "launch",
          "note": "Google now labels it a legacy model; not covered by the 3.6–3.8 introductory rate.",
          "source": "https://ai.google.dev/gemini-api/docs/changelog"
        }
      ],
      "tier": "mid"
    },
    "gemini-3.5-flash-lite": {
      "name": "Gemini 3.5 Flash-Lite",
      "provider": "gemini",
      "events": [
        {
          "date": "2026-07-21",
          "in": 0.3,
          "out": 2.5,
          "kind": "launch",
          "source": "https://ai.google.dev/gemini-api/docs/changelog"
        }
      ],
      "tier": "small"
    },
    "gemini-3.1-flash-lite": {
      "name": "Gemini 3.1 Flash-Lite",
      "provider": "gemini",
      "events": [
        {
          "date": "2026-05-07",
          "in": 0.25,
          "out": 1.5,
          "kind": "launch",
          "note": "Preview Mar 3, 2026. Google’s earliest shutdown date: May 7, 2027.",
          "source": "https://ai.google.dev/gemini-api/docs/changelog"
        }
      ],
      "tier": "small"
    },
    "deepseek-v4-pro": {
      "name": "DeepSeek V4 Pro",
      "provider": "deepseek",
      "events": [
        {
          "date": "2026-04-24",
          "in": 1.74,
          "out": 3.48,
          "kind": "launch",
          "confidence": "approx",
          "note": "V4 preview API opened Apr 24, 2026; price as recorded in June 2026 (DeepSeek publishes its table as an image).",
          "source": "https://api-docs.deepseek.com/news/news260424"
        },
        {
          "date": "2026-07",
          "in": 0.435,
          "out": 0.87,
          "kind": "cut",
          "confidence": "approx",
          "note": "About 75% off. Exact date not published — recorded by Jul 25, 2026."
        },
        {
          "date": "2026-08-16",
          "in": 1.32,
          "out": 3.96,
          "kind": "increase",
          "off_peak": {
            "in": 0.66,
            "out": 1.98
          },
          "note": "New peak / off-peak split from 16:00 UTC, Aug 16, 2026. Peak = Mon–Fri 09:00–12:00 and 14:00–18:00 Beijing time (01:00–04:00 and 06:00–10:00 UTC); off-peak is 50% lower. V4 Pro reached GA Aug 13.",
          "source": "https://api-docs.deepseek.com/news/news260813"
        }
      ],
      "polycog_notes": [
        {
          "date": "2026-08-18",
          "approx": false,
          "text": "PolyCog estimates at the peak rate around the clock (it never under-estimates); off-peak actuals settle in your provider bill."
        }
      ],
      "tier": "flagship"
    },
    "deepseek-v4-flash": {
      "name": "DeepSeek V4 Flash",
      "provider": "deepseek",
      "status": "legacy",
      "tier": "small",
      "events": [
        {
          "date": "2026-04-24",
          "in": 0.14,
          "out": 0.28,
          "kind": "launch",
          "confidence": "approx",
          "note": "Input price as recorded by PolyCog; output confirmed by press coverage of the August increase.",
          "source": "https://api-docs.deepseek.com/news/news260424"
        },
        {
          "date": "2026-08-16",
          "in": 0.44,
          "out": 1.32,
          "kind": "increase",
          "off_peak": {
            "in": 0.22,
            "out": 0.66
          },
          "note": "Same peak / off-peak split as V4 Pro.",
          "source": "https://api-docs.deepseek.com/news/news260813"
        },
        {
          "date": "2026-09-10",
          "in": 0.3,
          "out": 1.2,
          "kind": "cut",
          "off_peak": {
            "in": 0.15,
            "out": 0.6
          },
          "note": "DeepSeek-V4-Flash retired with the V4.1 Flash launch; the deepseek-v4-flash id is \"temporarily routed to V4.1 Flash\" and billed at the Flash price. PolyCog moved its catalog and sample tier to the new deepseek-flash id the same week; this row stays for sessions saved before the switch.",
          "source": "https://api-docs.deepseek.com/updates/"
        }
      ],
      "polycog_notes": [
        {
          "date": "2026-08-18",
          "approx": false,
          "text": "Estimated at the peak rate around the clock."
        }
      ],
      "retired": "2026-09-10",
      "note": "DeepSeek's V4-generation small model (Apr 24 - Sep 10, 2026): PolyCog's sample-tier shared DeepSeek model and cheapest DeepSeek pick until DeepSeek-V4.1-Flash replaced it under the new deepseek-flash id. The old id still answers as an alias of V4.1 Flash at the Flash rate."
    },
    "deepseek-flash": {
      "name": "DeepSeek V4.1 Flash",
      "provider": "deepseek",
      "tier": "small",
      "events": [
        {
          "date": "2026-09-10",
          "in": 0.3,
          "out": 1.2,
          "kind": "launch",
          "off_peak": {
            "in": 0.15,
            "out": 0.6
          },
          "note": "DeepSeek-V4.1-Flash under the new id deepseek-flash (\"Change the model name to `deepseek-flash` to call the latest V4.1 Flash model\"). Cache-hit input $0.006 peak / $0.003 off-peak; 1M context, 384K max output, native vision, thinking and non-thinking modes. New pricing effective 04:00 UTC, Sep 10, 2026; same peak / off-peak windows as V4 (Mon-Fri 01:00-04:00 and 06:00-10:00 UTC peak, 50% off otherwise).",
          "source": "https://api-docs.deepseek.com/news/news260910"
        }
      ],
      "polycog_notes": [
        {
          "date": "2026-09-11",
          "approx": false,
          "text": "Estimated at the peak rate around the clock (never under-estimates); off-peak actuals settle in your DeepSeek bill."
        }
      ]
    },
    "grok-4.6": {
      "name": "Grok 4.6",
      "provider": "xai",
      "events": [
        {
          "date": "2026-08-12",
          "in": 2,
          "out": 6,
          "kind": "launch",
          "note": "Prompts of 200K+ tokens bill $4 / $12 for the whole request.",
          "source": "https://x.ai/news/grok-4-6"
        }
      ],
      "tier": "flagship"
    },
    "grok-4.5": {
      "name": "Grok 4.5",
      "provider": "xai",
      "events": [
        {
          "date": "2026-07-16",
          "in": 2,
          "out": 6,
          "kind": "launch",
          "source": "https://x.ai/news/grok-4-5"
        }
      ],
      "tier": "flagship"
    },
    "grok-4.3": {
      "name": "Grok 4.3",
      "provider": "xai",
      "events": [
        {
          "date": "2026-05-01",
          "in": 1.25,
          "out": 2.5,
          "kind": "launch",
          "confidence": "secondary",
          "note": "Dated from launch coverage (Apr 30 / May 1, 2026); not on x.ai’s news index."
        }
      ],
      "tier": "flagship"
    },
    "davinci": {
      "name": "GPT-3 Davinci",
      "provider": "openai",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2021-09-01",
          "in": 60.0,
          "out": 60.0,
          "kind": "launch",
          "confidence": "approx",
          "source": "https://the-decoder.com/openai-cuts-prices-for-gpt-3-by-two-thirds/",
          "note": "Originally priced as $0.0600 per 1K tokens (single rate)."
        },
        {
          "date": "2022-09-01",
          "in": 20.0,
          "out": 20.0,
          "kind": "cut",
          "confidence": "secondary",
          "note": "Two-thirds price cut across the GPT-3 engine family, effective Sept 1 2022.",
          "source": "https://the-decoder.com/openai-cuts-prices-for-gpt-3-by-two-thirds/"
        }
      ],
      "retired": "2024-01-04",
      "note": "2,049-token context; largest of the original four GPT-3 base-model tiers, no instruction tuning or chat formatting."
    },
    "curie": {
      "name": "GPT-3 Curie",
      "provider": "openai",
      "status": "legacy",
      "tier": "mid",
      "events": [
        {
          "date": "2021-09-01",
          "in": 6.0,
          "out": 6.0,
          "kind": "launch",
          "confidence": "approx",
          "source": "https://the-decoder.com/openai-cuts-prices-for-gpt-3-by-two-thirds/",
          "note": "Originally priced as $0.0060 per 1K tokens (single rate)."
        },
        {
          "date": "2022-09-01",
          "in": 2.0,
          "out": 2.0,
          "kind": "cut",
          "confidence": "secondary",
          "note": "Two-thirds price cut across the GPT-3 engine family.",
          "source": "https://the-decoder.com/openai-cuts-prices-for-gpt-3-by-two-thirds/"
        }
      ],
      "retired": "2024-01-04",
      "note": "2,049-token context; second-largest GPT-3 base tier."
    },
    "babbage": {
      "name": "GPT-3 Babbage",
      "provider": "openai",
      "status": "legacy",
      "tier": "small",
      "events": [
        {
          "date": "2021-09-01",
          "in": 1.2,
          "out": 1.2,
          "kind": "launch",
          "confidence": "approx",
          "source": "https://the-decoder.com/openai-cuts-prices-for-gpt-3-by-two-thirds/",
          "note": "Originally priced as $0.0012 per 1K tokens (single rate)."
        },
        {
          "date": "2022-09-01",
          "in": 0.5,
          "out": 0.5,
          "kind": "cut",
          "confidence": "secondary",
          "note": "Two-thirds-class price cut across the GPT-3 engine family.",
          "source": "https://the-decoder.com/openai-cuts-prices-for-gpt-3-by-two-thirds/"
        }
      ],
      "retired": "2024-01-04",
      "note": "2,049-token context; third-tier GPT-3 base model."
    },
    "ada": {
      "name": "GPT-3 Ada",
      "provider": "openai",
      "status": "legacy",
      "tier": "small",
      "events": [
        {
          "date": "2021-09-01",
          "in": 0.8,
          "out": 0.8,
          "kind": "launch",
          "confidence": "approx",
          "source": "https://the-decoder.com/openai-cuts-prices-for-gpt-3-by-two-thirds/",
          "note": "Originally priced as $0.0008 per 1K tokens (single rate)."
        },
        {
          "date": "2022-09-01",
          "in": 0.4,
          "out": 0.4,
          "kind": "cut",
          "confidence": "secondary",
          "note": "Two-thirds-class price cut across the GPT-3 engine family.",
          "source": "https://the-decoder.com/openai-cuts-prices-for-gpt-3-by-two-thirds/"
        }
      ],
      "retired": "2024-01-04",
      "note": "2,049-token context; smallest/cheapest GPT-3 base model — OpenAI's cheapest generally-available text model from the 2021 price list until gpt-3.5-turbo-era models undercut it in 2024."
    },
    "text-davinci-003": {
      "name": "text-davinci-003",
      "provider": "openai",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2022-11-28",
          "in": 20.0,
          "out": 20.0,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://www.techzine.eu/news/devops/95696/openai-releases-new-gpt-3-generative-text-model/",
          "note": "Originally priced as $0.0200 per 1K tokens (single rate)."
        }
      ],
      "retired": "2024-01-04",
      "note": "4,097-token context; InstructGPT-tuned successor to the base GPT-3 engines, part of the \"GPT-3.5\" series OpenAI named at launch. Single per-1K-token rate (no split input/output pricing)."
    },
    "gpt-3.5-turbo": {
      "name": "GPT-3.5 Turbo",
      "provider": "openai",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2023-03-01",
          "in": 2.0,
          "out": 2.0,
          "kind": "launch",
          "source": "https://openai.com/index/introducing-chatgpt-and-whisper-apis/",
          "note": "Originally priced as $0.002 per 1K tokens (single blended rate)."
        },
        {
          "date": "2023-06-13",
          "in": 1.5,
          "out": 2.0,
          "kind": "cut",
          "note": "Split into separate input/output pricing for the first time (25% input cut); gpt-3.5-turbo-0613 and a 16K-context variant (gpt-3.5-turbo-16k, $3/$4 per 1M) also introduced same day.",
          "source": "https://openai.com/index/function-calling-and-other-api-updates/"
        },
        {
          "date": "2023-11-06",
          "in": 1.0,
          "out": 2.0,
          "kind": "cut",
          "note": "DevDay: gpt-3.5-turbo-1106 (16K context) cut input 3x vs the old 16K model.",
          "source": "https://openai.com/index/new-models-and-developer-products-announced-at-devday/"
        },
        {
          "date": "2024-01-25",
          "in": 0.5,
          "out": 1.5,
          "kind": "cut",
          "note": "gpt-3.5-turbo-0125 released: input cut 50%, output cut 25%. Became OpenAI's cheapest surviving model after the Jan 4 2024 GPT-3 shutdown left a brief gap.",
          "source": "https://openai.com/index/new-embedding-models-and-api-updates/"
        }
      ],
      "retired": "2026-10-23",
      "note": "4K-token context at launch (16K variant added Jun 2023). Powered the original ChatGPT API; was OpenAI's flagship/cheapest current-generation chat model for two weeks until GPT-4 launched, then settled into the mid/cheap "
    },
    "gpt-4": {
      "name": "GPT-4 (8K)",
      "provider": "openai",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2023-03-14",
          "in": 30.0,
          "out": 60.0,
          "kind": "launch",
          "source": "https://openai.com/index/gpt-4-research/",
          "note": "Originally priced as $0.03 per 1K prompt tokens / $0.06 per 1K completion tokens."
        }
      ],
      "retired": "2026-10-23",
      "note": "8,192-token context tier (a 32K tier, gpt-4-32k, was sold in parallel at 2x the price)."
    },
    "gpt-4-32k": {
      "name": "GPT-4 (32K)",
      "provider": "openai",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2023-03-14",
          "in": 60.0,
          "out": 120.0,
          "kind": "launch",
          "source": "https://openai.com/index/gpt-4-research/",
          "note": "Originally priced as $0.06 per 1K prompt tokens / $0.12 per 1K completion tokens."
        }
      ],
      "retired": "2025-06-06",
      "note": "32,768-token context tier of GPT-4, priced at 2x the 8K tier throughout its life."
    },
    "gpt-4-turbo": {
      "name": "GPT-4 Turbo",
      "provider": "openai",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2023-11-06",
          "in": 10.0,
          "out": 30.0,
          "kind": "launch",
          "source": "https://openai.com/index/new-models-and-developer-products-announced-at-devday/",
          "note": "Originally priced as $0.01 per 1K input tokens / $0.03 per 1K output tokens."
        }
      ],
      "retired": "2026-10-23",
      "note": "128K-token context, first frontier OpenAI model with a context window that large. Shipped first as a preview snapshot, later as a GA alias with vision merged in."
    },
    "gpt-4o": {
      "name": "GPT-4o",
      "provider": "openai",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2024-05-13",
          "in": 5.0,
          "out": 15.0,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://the-decoder.com/openai-cuts-gpt-4o-prices-and-quadruples-output-tokens/"
        },
        {
          "date": "2024-08-06",
          "in": 2.5,
          "out": 10.0,
          "kind": "cut",
          "confidence": "secondary",
          "note": "gpt-4o-2024-08-06 snapshot: 50% input cut, 33% output cut; output-token limit quadrupled same day.",
          "source": "https://the-decoder.com/openai-cuts-gpt-4o-prices-and-quadruples-output-tokens/"
        }
      ],
      "note": "128K-token context; first natively multimodal (text/audio/image) OpenAI flagship, replaced GPT-4 Turbo as default and cut price 50% vs it.",
      "last_checked": "2026-09-09"
    },
    "o1-preview": {
      "name": "o1-preview",
      "provider": "openai",
      "status": "legacy",
      "tier": "reasoning",
      "events": [
        {
          "date": "2024-09-12",
          "in": 15.0,
          "out": 60.0,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://venturebeat.com/programming-development/what-openais-new-o1-preview-and-o1-mini-models-mean-for-developers"
        }
      ],
      "retired": "2025-07-28",
      "note": "128K context; OpenAI's first publicly released reasoning ('thinking') model, previewing the o-series."
    },
    "o1-mini": {
      "name": "o1-mini",
      "provider": "openai",
      "status": "legacy",
      "tier": "reasoning",
      "events": [
        {
          "date": "2024-09-12",
          "in": 3.0,
          "out": 12.0,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://venturebeat.com/programming-development/what-openais-new-o1-preview-and-o1-mini-models-mean-for-developers"
        }
      ],
      "retired": "2025-10-27",
      "note": "128K context; cut-down/cheaper sibling of o1-preview, launched same day."
    },
    "o1": {
      "name": "o1",
      "provider": "openai",
      "status": "legacy",
      "tier": "reasoning",
      "events": [
        {
          "date": "2024-12-17",
          "in": 15.0,
          "out": 60.0,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://openrouter.ai/openai/o1"
        }
      ],
      "retired": "2026-10-23",
      "note": "200K context at GA. Full general-availability release of the o1 reasoning model, promoted from o1-preview at the same price."
    },
    "o3-mini": {
      "name": "o3-mini",
      "provider": "openai",
      "status": "legacy",
      "tier": "reasoning",
      "events": [
        {
          "date": "2025-01-31",
          "in": 1.1,
          "out": 4.4,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://simonwillison.net/2025/Jan/31/o3-mini/"
        }
      ],
      "retired": "2026-10-23",
      "note": "200K context; cost-efficient reasoning model, initially gated to Tier 3+ API accounts."
    },
    "o3": {
      "name": "o3",
      "provider": "openai",
      "status": "legacy",
      "tier": "reasoning",
      "events": [
        {
          "date": "2025-04-16",
          "in": 10.0,
          "out": 40.0,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://venturebeat.com/ai/openai-announces-80-price-drop-for-o3-its-most-powerful-reasoning-model"
        },
        {
          "date": "2025-06-10",
          "in": 2.0,
          "out": 8.0,
          "kind": "cut",
          "note": "80% price cut across the board; OpenAI framed it as passing through efficiency gains, alongside the launch of o3-pro.",
          "source": "https://simonwillison.net/2025/Jun/10/o3-price-drop/"
        }
      ],
      "retired": "2026-12-11",
      "note": "200K context; OpenAI's flagship reasoning model through mid-2025, until GPT-5 unified reasoning and chat into one model."
    },
    "o4-mini": {
      "name": "o4-mini",
      "provider": "openai",
      "status": "legacy",
      "tier": "reasoning",
      "events": [
        {
          "date": "2025-04-16",
          "in": 1.1,
          "out": 4.4,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://www.datacamp.com/blog/o4-mini"
        }
      ],
      "retired": "2026-10-23",
      "note": "200K context; launched alongside o3 as the cost-efficient reasoning option, succeeding o3-mini."
    },
    "gpt-4.5-preview": {
      "name": "GPT-4.5 (research preview)",
      "provider": "openai",
      "status": "legacy",
      "tier": "other",
      "events": [
        {
          "date": "2025-02-27",
          "in": 75.0,
          "out": 150.0,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://simonwillison.net/2025/Feb/27/introducing-gpt-45/"
        }
      ],
      "retired": "2025-07-14",
      "note": "128K context; OpenAI's largest non-reasoning base model, a short-lived and very expensive research preview."
    },
    "gpt-4.1-nano": {
      "name": "GPT-4.1 nano",
      "provider": "openai",
      "status": "legacy",
      "tier": "small",
      "events": [
        {
          "date": "2025-04-14",
          "in": 0.1,
          "out": 0.4,
          "kind": "launch",
          "source": "https://simonwillison.net/2025/Apr/14/gpt-4-1/"
        }
      ],
      "note": "1,000,000-token context; OpenAI's cheapest model at launch, undercutting gpt-4o-mini.",
      "last_checked": "2026-09-09"
    },
    "gpt-5": {
      "name": "GPT-5",
      "provider": "openai",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2025-08-07",
          "in": 1.25,
          "out": 10.0,
          "kind": "launch",
          "source": "https://openai.com/index/introducing-gpt-5/"
        }
      ],
      "retired": "2026-12-11",
      "note": "400K-token context. Unified OpenAI's reasoning and non-reasoning lines into a single default model, replacing GPT-4o, o3, o4-mini, GPT-4.1 and GPT-4.5 as the ChatGPT default."
    },
    "gpt-5-mini": {
      "name": "GPT-5 mini",
      "provider": "openai",
      "status": "legacy",
      "tier": "mid",
      "events": [
        {
          "date": "2025-08-07",
          "in": 0.25,
          "out": 2.0,
          "kind": "launch",
          "source": "https://openai.com/index/introducing-gpt-5/"
        }
      ],
      "note": "400K-token context; mid-cost member of the GPT-5 family.",
      "last_checked": "2026-09-09"
    },
    "gpt-5-nano": {
      "name": "GPT-5 nano",
      "provider": "openai",
      "status": "legacy",
      "tier": "small",
      "events": [
        {
          "date": "2025-08-07",
          "in": 0.05,
          "out": 0.4,
          "kind": "launch",
          "source": "https://openai.com/index/introducing-gpt-5/"
        }
      ],
      "note": "400K-token context; cheapest generally-available OpenAI text model from launch through at least Sept 2026 (undercut by no later nano/mini variant found in research).",
      "last_checked": "2026-09-09"
    },
    "gpt-5.1-nano": {
      "name": "GPT-5.1 nano",
      "provider": "openai",
      "status": "legacy",
      "tier": "small",
      "events": [
        {
          "date": "2025-11-13",
          "in": 0.05,
          "out": 0.4,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://langcopilot.com/gpt-5-1-nano-token-calculator"
        }
      ],
      "note": "Small member of the GPT-5.1 family, priced identically to gpt-5-nano (no change to the cheapest-tier price point).",
      "last_checked": "2026-09-09"
    },
    "gpt-5.2": {
      "name": "GPT-5.2",
      "provider": "openai",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2025-12-11",
          "in": 1.75,
          "out": 14.0,
          "kind": "launch",
          "source": "https://openai.com/index/introducing-gpt-5-2/"
        }
      ],
      "note": "Released amid competitive pressure from Google Gemini 3 (\"code red\" reporting). A Pro tier (gpt-5.2-pro, $21/$168) also shipped. OpenAI reiterated no plans to deprecate GPT-5.1, GPT-5 or GPT-4.1.",
      "last_checked": "2026-09-09"
    },
    "gpt-5.3-codex": {
      "name": "GPT-5.3-Codex",
      "provider": "openai",
      "status": "legacy",
      "tier": "other",
      "events": [
        {
          "date": "2026-01-01",
          "in": 1.75,
          "out": 14.0,
          "kind": "launch",
          "confidence": "approx",
          "source": "https://developers.openai.com/api/docs/pricing"
        }
      ],
      "note": "Coding-specialised model; no general-purpose gpt-5.3 flagship was released — the flagship line went 5.2 → 5.4 directly, with 5.3 shipping only as this Codex-specific variant.",
      "last_checked": "2026-09-09"
    },
    "gpt-5.4-nano": {
      "name": "GPT-5.4 nano",
      "provider": "openai",
      "status": "legacy",
      "tier": "small",
      "events": [
        {
          "date": "2026-03-17",
          "in": 0.2,
          "out": 1.25,
          "kind": "launch",
          "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano/"
        }
      ],
      "note": "Smallest member of the GPT-5.4 family; priced above gpt-5-nano/gpt-5.1-nano, so it did not become the new cheapest model.",
      "last_checked": "2026-09-09"
    },
    "claude-1": {
      "name": "Claude (v1: claude-1.0 / 1.1 / 1.2 / 1.3)",
      "provider": "anthropic",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2023-03-14",
          "in": 11.02,
          "out": 32.68,
          "kind": "launch",
          "confidence": "approx",
          "source": "https://www-cdn.anthropic.com/b0d46f8c8e2f4cca2833c17cb68435035c0bb5a9/apr-pricing-tokens_2023-05-10-213025_tnty.pdf",
          "note": "Originally priced as $11.02 / $32.68 per million tokens (Prompt / Completion) — figure verified from Anthropic's pricing PDF dated 2023-05-10."
        }
      ],
      "retired": "2024-11-06",
      "note": "9K context at launch; expanded to 100K tokens on 2023-05-11. API ids claude-1.0, claude-1.1, claude-1.2, claude-1.3 (also exposed as claude-v1 / claude-v1.3). Superseded by Claude 2 on 2023-07-11 but remained accessible "
    },
    "claude-instant-1": {
      "name": "Claude Instant (1.0 / 1.1 / 1.2)",
      "provider": "anthropic",
      "status": "legacy",
      "tier": "small",
      "events": [
        {
          "date": "2023-03-14",
          "in": 1.63,
          "out": 5.51,
          "kind": "launch",
          "confidence": "approx",
          "source": "https://www-cdn.anthropic.com/b0d46f8c8e2f4cca2833c17cb68435035c0bb5a9/apr-pricing-tokens_2023-05-10-213025_tnty.pdf",
          "note": "Originally priced as $1.63 / $5.51 per million tokens (Prompt / Completion) — verified from Anthropic's pricing PDF dated 2023-05-10; launch-."
        },
        {
          "date": "2023-12-13",
          "in": 0.8,
          "out": 2.4,
          "kind": "cut",
          "confidence": "approx",
          "note": "Cut to $0.80/$2.40 (-51%/-56%). Date bracketed: Anthropic's Nov 2023 pricing PDF (which already lists Claude 2.1, so >= 2023-11-21) still shows Instant at $1.63/$5.51; the updated Anthropic PDF shows $0.80/$2.40 and a Hacker News post noticed the change on 202",
          "source": "https://www-cdn.anthropic.com/31021aea87c30ccaecbd2e966e49a03834bfd1d2/pricing.pdf"
        }
      ],
      "retired": "2024-11-06",
      "note": "9K context at launch, 100K from 2023-05-11. Claude Instant 1.2 (claude-instant-1.2) released 2023-08-09 at the unchanged $1.63/$5.51 (Nov 2023 PDF still shows $1.63/$5.51). API ids claude-instant-1.0, claude-instant-1.1,"
    },
    "claude-2.0": {
      "name": "Claude 2",
      "provider": "anthropic",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2023-07-11",
          "in": 11.02,
          "out": 32.68,
          "kind": "launch",
          "source": "https://www-cdn.anthropic.com/90df03aed08b794ab03c5a7bf28b2ad9cf26cf3c/model_pricing_july2023.pdf"
        },
        {
          "date": "2023-11-21",
          "in": 8.0,
          "out": 24.0,
          "kind": "cut",
          "note": "Cut to $8/$24 (-27%) alongside the Claude 2.1 launch (\"We are also updating our pricing to improve cost efficiency for our customers across models\"). Anthropic's Nov 2023 pricing PDF lists Claude 2.0 and 2.1 both at $8.00/$24.00.",
          "source": "https://www-cdn.anthropic.com/1b1ea2c43d8dd058f6a331a8097e05ea40d626c6/model_pricing_nov2023.pdf"
        }
      ],
      "retired": "2025-07-21",
      "note": "100K token context. API id claude-2.0 (initially claude-2). Launched at the same price as Claude 1.3; cut to $8/$24 with the Claude 2.1 launch."
    },
    "claude-2.1": {
      "name": "Claude 2.1",
      "provider": "anthropic",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2023-11-21",
          "in": 8.0,
          "out": 24.0,
          "kind": "launch",
          "source": "https://www-cdn.anthropic.com/1b1ea2c43d8dd058f6a331a8097e05ea40d626c6/model_pricing_nov2023.pdf"
        }
      ],
      "retired": "2025-07-21",
      "note": "200K token context (doubled from Claude 2's 100K); 2x fewer false statements vs Claude 2.0; tool use beta. Same $8/$24 price as the simultaneously cut Claude 2.0."
    },
    "claude-3-opus-20240229": {
      "name": "Claude 3 Opus",
      "provider": "anthropic",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2024-03-04",
          "in": 15.0,
          "out": 75.0,
          "kind": "launch",
          "source": "https://www.anthropic.com/news/claude-3-family"
        }
      ],
      "retired": "2026-01-05",
      "note": "200K context. Price never changed ($15/$75) from launch to retirement. Superseded as Anthropic's 'most intelligent' model by Claude 3.5 Sonnet (2024-06-20) at 1/5 the price."
    },
    "claude-3-sonnet-20240229": {
      "name": "Claude 3 Sonnet",
      "provider": "anthropic",
      "status": "legacy",
      "tier": "mid",
      "events": [
        {
          "date": "2024-03-04",
          "in": 3.0,
          "out": 15.0,
          "kind": "launch",
          "source": "https://www.anthropic.com/news/claude-3-family"
        }
      ],
      "retired": "2025-07-21",
      "note": "200K context. Established the $3/$15 Sonnet price point that held through Sonnet 4.6. Replaced by Claude 3.5 Sonnet (2024-06-20)."
    },
    "claude-3-haiku-20240307": {
      "name": "Claude 3 Haiku",
      "provider": "anthropic",
      "status": "legacy",
      "tier": "small",
      "events": [
        {
          "date": "2024-03-13",
          "in": 0.25,
          "out": 1.25,
          "kind": "launch",
          "source": "https://www.anthropic.com/news/claude-3-haiku"
        }
      ],
      "retired": "2026-04-20",
      "note": "200K context. Announced 2024-03-04 (\"Haiku will be available soon\"), released via API 2024-03-13. Cheapest Anthropic model from release until its retirement on 2026-04-20 (Claude 3.5 Haiku launched at 4x its price and ev"
    },
    "claude-3-5-sonnet-20240620": {
      "name": "Claude 3.5 Sonnet",
      "provider": "anthropic",
      "status": "legacy",
      "tier": "mid",
      "events": [
        {
          "date": "2024-06-20",
          "in": 3.0,
          "out": 15.0,
          "kind": "launch",
          "source": "https://www.anthropic.com/news/claude-3-5-sonnet"
        }
      ],
      "retired": "2025-10-28",
      "note": "200K context. Positioned by Anthropic as its most intelligent model at launch (outperforming Claude 3 Opus at 1/5 the cost, 2x speed). The launch post is stamped 'Jun 21, 2024' in some timezones; the model id and press d"
    },
    "claude-3-5-sonnet-20241022": {
      "name": "Claude 3.5 Sonnet (upgraded, Oct 2024)",
      "provider": "anthropic",
      "status": "legacy",
      "tier": "mid",
      "events": [
        {
          "date": "2024-10-22",
          "in": 3.0,
          "out": 15.0,
          "kind": "launch",
          "source": "https://www.anthropic.com/news/3-5-models-and-computer-use"
        }
      ],
      "retired": "2025-10-28",
      "note": "200K context. Same price and speed as the June 2024 3.5 Sonnet; introduced the computer-use beta. Both 3.5 Sonnet snapshots were deprecated and retired together."
    },
    "claude-3-5-haiku-20241022": {
      "name": "Claude 3.5 Haiku",
      "provider": "anthropic",
      "status": "legacy",
      "tier": "small",
      "events": [
        {
          "date": "2024-11-04",
          "in": 1.0,
          "out": 5.0,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://techcrunch.com/2024/11/04/anthropic-hikes-the-price-of-its-haiku-model"
        },
        {
          "date": "2024-12-03",
          "in": 0.8,
          "out": 4.0,
          "kind": "cut",
          "note": "Cut from $1/$5 to $0.80/$4 (-20%) one month after launch; Anthropic appended a dated update to the October announcement post.",
          "source": "https://www.anthropic.com/news/3-5-models-and-computer-use"
        }
      ],
      "retired": "2026-02-19",
      "note": "200K context; text-only at launch, vision added 2025-02-24. Announced 2024-10-22 as coming 'later this month' at the same cost as Claude 3 Haiku ($0.25/$1.25, per contemporaneous press); actually released 2024-11-04 at $"
    },
    "claude-3-7-sonnet-20250219": {
      "name": "Claude 3.7 Sonnet",
      "provider": "anthropic",
      "status": "legacy",
      "tier": "mid",
      "events": [
        {
          "date": "2025-02-24",
          "in": 3.0,
          "out": 15.0,
          "kind": "launch",
          "source": "https://www.anthropic.com/news/claude-3-7-sonnet"
        }
      ],
      "retired": "2026-02-19",
      "note": "200K context; first hybrid-reasoning Claude (extended thinking, thinking tokens billed as output; 128K output beta). Same $3/$15 as prior Sonnets. Launched with Claude Code research preview."
    },
    "claude-opus-4-20250514": {
      "name": "Claude Opus 4",
      "provider": "anthropic",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2025-05-22",
          "in": 15.0,
          "out": 75.0,
          "kind": "launch",
          "source": "https://www.anthropic.com/news/claude-4"
        }
      ],
      "retired": "2026-06-15",
      "note": "200K context, 32K max output, extended thinking. Restored the $15/$75 Opus price point after the 3.5/3.7 Sonnet era. Replaced by Opus 4.1 (2025-08-05)."
    },
    "claude-sonnet-4-20250514": {
      "name": "Claude Sonnet 4",
      "provider": "anthropic",
      "status": "legacy",
      "tier": "mid",
      "events": [
        {
          "date": "2025-05-22",
          "in": 3.0,
          "out": 15.0,
          "kind": "launch",
          "source": "https://www.anthropic.com/news/claude-4"
        }
      ],
      "retired": "2026-06-15",
      "note": "200K context (1M beta from 2025-08-12 with a long-context premium of $6/$22.50 for prompts >200K input tokens — not a base-tier change). Base price $3/$15 unchanged for life. Replaced by Sonnet 4.5 (2025-09-29)."
    },
    "claude-opus-4-1-20250805": {
      "name": "Claude Opus 4.1",
      "provider": "anthropic",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2025-08-05",
          "in": 15.0,
          "out": 75.0,
          "kind": "launch",
          "source": "https://www.anthropic.com/news/claude-opus-4-1"
        }
      ],
      "retired": "2026-08-05",
      "note": "200K context. Incremental update to Opus 4 at the same $15/$75; the last $15/$75 Opus. Stayed at $15/$75 after Opus 4.5 launched at $5/$25 (pricing page still lists it at $15/$75 as retired). Replacement per deprecation "
    },
    "claude-sonnet-4-5-20250929": {
      "name": "Claude Sonnet 4.5",
      "provider": "anthropic",
      "status": "legacy",
      "tier": "mid",
      "events": [
        {
          "date": "2025-09-29",
          "in": 3.0,
          "out": 15.0,
          "kind": "launch",
          "source": "https://www.anthropic.com/news/claude-sonnet-4-5"
        }
      ],
      "note": "200K context (1M beta with >200K premium, same as Sonnet 4). Anthropic called it 'the best model in the world for agents, coding, and computer use' with 'the highest intelligence across most tasks' — i.e. positioned abov",
      "last_checked": "2026-09-09"
    },
    "claude-opus-4-5-20251101": {
      "name": "Claude Opus 4.5",
      "provider": "anthropic",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2025-11-24",
          "in": 5.0,
          "out": 25.0,
          "kind": "launch",
          "source": "https://www.anthropic.com/news/claude-opus-4-5"
        }
      ],
      "note": "200K context. The Opus price cut: $15/$75 -> $5/$25 (-67%), the price point every later Opus (4.6, 4.7, 4.8, 5) kept. No pricing or model events occurred between Opus 4.5 (2025-11-24) and Opus 4.6 (2026-02-05) other than",
      "last_checked": "2026-09-09"
    },
    "gemini-1.0-pro": {
      "name": "Gemini 1.0 Pro (Gemini Pro)",
      "provider": "gemini",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2024-02-15",
          "in": 0.5,
          "out": 1.5,
          "kind": "launch",
          "confidence": "approx",
          "source": "https://techcrunch.com/2024/02/15/google-makes-more-gemini-models-available-to-developers-but-needs-to-work-on-its-branding",
          "note": "Originally priced as $0.000125 per 1,000 characters (input) / $0.000375 per 1,000 characters (output) — converted to per-1,000,000-tokens at ."
        }
      ],
      "retired": "2025-02-18",
      "note": "~32K token context (30,720 input / 2,048 output). Text-only 'Gemini Pro'; a multimodal 'Gemini Pro Vision' variant launched alongside at the same per-unit price. Free preview from 2023-12-13; priced GA from 2024-02-15; b"
    },
    "gemini-1.5-pro": {
      "name": "Gemini 1.5 Pro",
      "provider": "gemini",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2024-05-23",
          "in": 3.5,
          "out": 10.5,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://news.ycombinator.com/item?id=41638568",
          "note": "Originally priced as $3.50 / $10.50 per 1,000,000 tokens for prompts <=128K tokens."
        },
        {
          "date": "2024-10-01",
          "in": 1.25,
          "out": 5.0,
          "kind": "cut",
          "note": "64% cut on input tokens, 52% cut on output tokens (Google's own percentages) for prompts <=128K, announced 2024-09-24 and effective 2024-10-01, alongside the gemini-1.5-pro-002 model refresh and higher (1,000 RPM) paid-tier rate limits.",
          "source": "https://developers.googleblog.com/en/updated-gemini-models-reduced-15-pro-pricing-increased-rate-limits-and-more/"
        }
      ],
      "retired": "2025-09-29",
      "note": "1M token context (2M in limited preview from 2024-08); long-context premium tier existed above 128K (roughly 2x the base price). Preview released 2024-04-09 (gemini-1.5-pro-latest, limited/free); GA with real pay-as-you-"
    },
    "gemini-1.5-flash": {
      "name": "Gemini 1.5 Flash",
      "provider": "gemini",
      "status": "legacy",
      "tier": "mid",
      "events": [
        {
          "date": "2024-05-23",
          "in": 0.35,
          "out": 1.05,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://www.neowin.net/news/google-slashes-gemini-15-flash-prices-igniting-llm-price-war/",
          "note": "Originally priced as $0.35 / $1.05 per 1,000,000 tokens for prompts <=128K tokens."
        },
        {
          "date": "2024-08-12",
          "in": 0.075,
          "out": 0.3,
          "kind": "cut",
          "note": "78% cut on input, 71% cut on output, for prompts <=128K tokens, with comparable cuts to the >128K tier and to context caching.",
          "source": "https://developers.googleblog.com/en/gemini-15-flash-updates-google-ai-studio-gemini-api/"
        }
      ],
      "retired": "2025-09-29",
      "note": "1M token context. Google's first 'Flash' fast/cheap general-purpose tier, distilled from 1.5 Pro. Preview released 2024-05-10 (gemini-1.5-flash-latest); GA 2024-05-23 (gemini-1.5-flash-001); refreshed gemini-1.5-flash-00"
    },
    "gemini-1.5-flash-8b": {
      "name": "Gemini 1.5 Flash-8B",
      "provider": "gemini",
      "status": "legacy",
      "tier": "small",
      "events": [
        {
          "date": "2024-10-03",
          "in": 0.0375,
          "out": 0.15,
          "kind": "launch",
          "source": "https://developers.googleblog.com/en/gemini-15-flash-8b-is-now-generally-available-for-use/",
          "note": "Originally priced as $0.0375 / $0.15 per 1,000,000 tokens for prompts <=128K tokens."
        }
      ],
      "retired": "2025-09-29",
      "note": "1M token context. Distilled 8B-parameter version of 1.5 Flash for lower latency/cost at high volume; announced in preview 2024-08-08 alongside the 1.5 Flash price cut, reached GA as gemini-1.5-flash-8b-001 on 2024-10-03."
    },
    "gemini-2.0-flash": {
      "name": "Gemini 2.0 Flash",
      "provider": "gemini",
      "status": "legacy",
      "tier": "mid",
      "events": [
        {
          "date": "2025-02-05",
          "in": 0.1,
          "out": 0.4,
          "kind": "launch",
          "source": "https://developers.googleblog.com/en/start-building-with-the-gemini-2-0-flash-family/",
          "note": "Originally priced as $0.10 / $0.40 per 1,000,000 tokens (single tier, no long-context split)."
        }
      ],
      "retired": "2026-06-01",
      "note": "1M token context. Experimental preview released free 2024-12-11 (gemini-2.0-flash-exp); GA 2025-02-05 as gemini-2.0-flash-001, at a single flat per-token price regardless of prompt length (dropping the 1.5 Flash <=128K/>"
    },
    "gemini-2.0-flash-lite": {
      "name": "Gemini 2.0 Flash-Lite",
      "provider": "gemini",
      "status": "legacy",
      "tier": "small",
      "events": [
        {
          "date": "2025-02-25",
          "in": 0.075,
          "out": 0.3,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://pricepertoken.com/pricing-page/model/google-gemini-2.0-flash-lite-001",
          "note": "Originally priced as $0.075 / $0.30 per 1,000,000 tokens (single tier)."
        }
      ],
      "retired": "2026-06-01",
      "note": "1M token context. Public preview announced 2025-02-05 alongside 2.0 Flash's GA; reached GA 2025-02-25. Also a single flat per-token price (no long-context split). Became the overall cheapest GA Gemini model on 2025-09-29"
    },
    "gemini-2.5-pro": {
      "name": "Gemini 2.5 Pro",
      "provider": "gemini",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2025-04-04",
          "in": 1.25,
          "out": 10.0,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://techcrunch.com/2025/04/04/gemini-2-5-pro-is-googles-most-expensive-ai-model-yet/",
          "note": "Originally priced as $1.25 / $10.00 per 1,000,000 tokens for prompts <=200K tokens ($2.50 / $15.00 above 200K)."
        },
        {
          "date": "2025-06-17",
          "in": 1.25,
          "out": 10.0,
          "kind": "permanent",
          "note": "Reached general availability (plain 'gemini-2.5-pro' id) at unchanged pricing from the paid preview.",
          "source": "https://cloud.google.com/blog/products/ai-machine-learning/gemini-2-5-flash-lite-flash-pro-ga-vertex-ai"
        }
      ],
      "note": "1M token context; long-context premium above 200K ($2.50/$15). 'Adaptive thinking' reasoning model. Free experimental preview from ~2025-03-25; first PAID public preview (gemini-2.5-pro-preview-03-25, then -05-06, -06-05",
      "last_checked": "2026-09-09"
    },
    "gemini-2.5-flash": {
      "name": "Gemini 2.5 Flash",
      "provider": "gemini",
      "status": "legacy",
      "tier": "mid",
      "events": [
        {
          "date": "2025-04-17",
          "in": 0.15,
          "out": 0.6,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://www.bgr.com/tech/gemini-2-5-flash-is-googles-cheapest-thinking-ai-what-you-need-to-know/",
          "note": "Originally priced as $0.15 / $0.60 per 1,000,000 tokens (non-thinking output; output priced at $3.50/1M when the thinking budget is used)."
        },
        {
          "date": "2025-06-17",
          "in": 0.3,
          "out": 2.5,
          "kind": "increase",
          "confidence": "approx",
          "note": "At GA, pricing was revised upward to a single unified rate (both thinking and non-thinking output billed the same), replacing the preview's cheap non-thinking-output / $3.50-thinking split. Net effect: 2x higher input, higher effective output for non-thinking ",
          "source": "https://ai.google.dev/gemini-api/docs/pricing"
        }
      ],
      "note": "1M token context. Hybrid reasoning ('thinking') model with a configurable thinking budget (0-24,576 tokens). Paid preview (gemini-2.5-flash-preview-04-17) launched 2025-04-17 with cheap non-thinking output but a much hig",
      "last_checked": "2026-09-09"
    },
    "gemini-3-pro-preview": {
      "name": "Gemini 3 Pro Preview",
      "provider": "gemini",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2025-11-18",
          "in": 2.0,
          "out": 12.0,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://venturebeat.com/technology/google-launches-gemini-3-1-pro-retaking-ai-crown-with-2x-reasoning",
          "note": "Originally priced as $2.00 / $12.00 per 1,000,000 tokens for prompts <=200K tokens ($4.00 / $18.00 above 200K)."
        }
      ],
      "retired": "2026-03-09",
      "note": "1M token context (long-context premium above 200K: $4/$18). 'State-of-the-art reasoning and multimodal understanding' flagship, succeeding Gemini 2.5 Pro. Launched 2025-11-18. Deprecated 2026-02-26 and shut down 2026-03-"
    },
    "deepseek-chat": {
      "name": "DeepSeek Chat (V2 → V2.5 → V3 → V3.1 → V3.2)",
      "provider": "deepseek",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2024-05-06",
          "in": 0.14,
          "out": 0.28,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://github.com/deepseek-ai/DeepSeek-V2",
          "note": "Originally priced as ¥1 per 1M input tokens / ¥2 per 1M output tokens (no caching tier existed yet)."
        },
        {
          "date": "2024-12-26",
          "in": 0.14,
          "out": 0.28,
          "kind": "promo",
          "note": "deepseek-chat upgraded in place to DeepSeek-V3; DeepSeek explicitly kept the same rate as V2 through a promotional window running to 2025-02-08.",
          "source": "https://www.deepseek.com/en/news/deepseek-v3/",
          "until": "2025-02-08"
        },
        {
          "date": "2025-02-09",
          "in": 0.27,
          "out": 1.1,
          "kind": "increase",
          "confidence": "secondary",
          "note": "Promotional V3 pricing ended (45-day promo begun 2024-12-26); cache-hit input rose to $0.07/1M. Press reported the output figure as $1.09-$1.10/1M depending on rounding of the underlying ¥8 rate.",
          "source": "https://technode.com/2025/02/10/deepseek-v3-ends-promotional-pricing-updates-api-service-rates/"
        },
        {
          "date": "2025-02-26",
          "in": 0.27,
          "out": 1.1,
          "kind": "promo",
          "note": "Off-peak discount program begins: DeepSeek-V3 (deepseek-chat) at 50% off, daily 16:30-00:30 UTC. Standard/peak rate unchanged. Program ran until 2025-09-05.",
          "source": "https://x.com/deepseek_ai/status/1894710448676884671",
          "until": "2025-09-05"
        },
        {
          "date": "2025-09-05",
          "in": 0.56,
          "out": 1.68,
          "kind": "increase",
          "note": "DeepSeek-V3.1 (hybrid thinking/non-thinking architecture) launched 2025-08-21; unified pricing for deepseek-chat and deepseek-reasoner took effect and the Feb 2025 off-peak discount program ended, both at 16:00 UTC on this date. Cache-hit input stayed $0.07/1M",
          "source": "https://api-docs.deepseek.com/news/news250821/"
        },
        {
          "date": "2025-09-29",
          "in": 0.28,
          "out": 0.42,
          "kind": "cut",
          "note": "DeepSeek-V3.2-Exp launched with DeepSeek Sparse Attention; cache-hit input fell to $0.028/1M. Described by DeepSeek as a 50%+ reduction from V3.1/V3.1-Terminus pricing.",
          "source": "https://api-docs.deepseek.com/news/news250929/"
        },
        {
          "date": "2026-04-24",
          "in": 0.14,
          "out": 0.28,
          "kind": "cut",
          "note": "DeepSeek-V4 launched; the legacy 'deepseek-chat' id was temporarily remapped to DeepSeek-V4-Flash's non-thinking mode at V4-Flash's launch price, pending full retirement.",
          "source": "https://api-docs.deepseek.com/news/news260424/"
        }
      ],
      "retired": "2026-07-24",
      "note": "Persistent API id for DeepSeek's general-purpose non-reasoning model. 128K context under V2/V2.5; DeepSeek's API exposed a 64K context window for V3 despite the checkpoint supporting 128K (see deepseek-ai/DeepSeek-V3 Git"
    },
    "deepseek-reasoner": {
      "name": "DeepSeek Reasoner (R1 → R1-0528 → V3.1/V3.2 thinking mode)",
      "provider": "deepseek",
      "status": "legacy",
      "tier": "reasoning",
      "events": [
        {
          "date": "2025-01-20",
          "in": 0.55,
          "out": 2.19,
          "kind": "launch",
          "source": "https://x.com/deepseek_ai/status/1881318145761439995"
        },
        {
          "date": "2025-02-26",
          "in": 0.55,
          "out": 2.19,
          "kind": "promo",
          "note": "Off-peak discount program begins: DeepSeek-R1 at 75% off, daily 16:30-00:30 UTC. Standard/peak rate unchanged. Program ran until 2025-09-05.",
          "source": "https://x.com/deepseek_ai/status/1894710448676884671",
          "until": "2025-09-05"
        },
        {
          "date": "2025-09-05",
          "in": 0.56,
          "out": 1.68,
          "kind": "cut",
          "note": "Unified with deepseek-chat under the V3.1 hybrid architecture (launched 2025-08-21) and the off-peak discount program ended, both at 16:00 UTC. Cache-hit input halved ($0.14→$0.07); cache-miss input rose $0.01 (effectively flat, $0.55→$0.56); output fell ~23% ",
          "source": "https://api-docs.deepseek.com/news/news250821/"
        },
        {
          "date": "2025-09-29",
          "in": 0.28,
          "out": 0.42,
          "kind": "cut",
          "note": "DeepSeek-V3.2-Exp launched; deepseek-reasoner (thinking mode) priced identically to deepseek-chat.",
          "source": "https://api-docs.deepseek.com/news/news250929/"
        },
        {
          "date": "2026-04-24",
          "in": 0.14,
          "out": 0.28,
          "kind": "cut",
          "note": "DeepSeek-V4 launched; the legacy 'deepseek-reasoner' id was temporarily remapped to DeepSeek-V4-Flash's thinking mode at V4-Flash's launch price, pending full retirement.",
          "source": "https://api-docs.deepseek.com/news/news260424/"
        }
      ],
      "retired": "2026-07-24",
      "note": "Chain-of-thought 'thinking mode' id, introduced as DeepSeek-R1. 64K context at launch; reasoning tokens are billed as output tokens. Merged into the V3.1 hybrid architecture (128K context) as its thinking-mode endpoint f"
    },
    "grok-beta": {
      "name": "Grok Beta",
      "provider": "xai",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2024-10-21",
          "in": 5.0,
          "out": 15.0,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://techcrunch.com/2024/10/21/xai-elon-musks-ai-startup-launches-an-api"
        },
        {
          "date": "2024-11-04",
          "in": 5.0,
          "out": 15.0,
          "kind": "promo",
          "note": "xAI's official 'API Public Beta' post reconfirmed grok-beta's price/128K context and introduced $25/month in free API credits for every account (prepaid holders got matching credits, e.g. a $50 top-up yielded $75/month), running through the end of December 202",
          "source": "https://x.ai/news/api"
        }
      ],
      "note": "128,000-token context (text-only). A companion multimodal grok-vision-beta shipped about a week later and is excluded here as a vision model. grok-beta was xAI's only API model for its first ~7 weeks.",
      "last_checked": "2026-09-09"
    },
    "grok-2-1212": {
      "name": "Grok 2 (1212)",
      "provider": "xai",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2024-12-14",
          "in": 2.0,
          "out": 10.0,
          "kind": "launch",
          "source": "https://x.com/xai/status/1868045132760842734"
        }
      ],
      "note": "128,000-token context. A vision sibling, grok-2-vision-1212 (32K context), launched the same day and is excluded here. Replaced grok-beta as xAI's flagship model, bundling a model upgrade with a price cut.",
      "last_checked": "2026-09-09"
    },
    "grok-3": {
      "name": "Grok 3",
      "provider": "xai",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2025-04-09",
          "in": 3.0,
          "out": 15.0,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://techcrunch.com/2025/04/09/elon-musks-ai-company-xai-launches-an-api-for-grok-3"
        }
      ],
      "retired": "2026-05-15",
      "note": "131,072-token (~128K) context — short of the 1M-token capacity xAI had claimed for the underlying model. The chat model was unveiled Feb 17, 2025, but API pricing/access followed about 7 weeks later."
    },
    "grok-3-fast": {
      "name": "Grok 3 Fast",
      "provider": "xai",
      "status": "legacy",
      "tier": "mid",
      "events": [
        {
          "date": "2025-04-09",
          "in": 5.0,
          "out": 25.0,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://techcrunch.com/2025/04/09/elon-musks-ai-company-xai-launches-an-api-for-grok-3"
        }
      ],
      "note": "Same 131K context and capability as grok-3, offered at a premium for lower latency.",
      "last_checked": "2026-09-09"
    },
    "grok-3-mini": {
      "name": "Grok 3 Mini",
      "provider": "xai",
      "status": "legacy",
      "tier": "reasoning",
      "events": [
        {
          "date": "2025-04-09",
          "in": 0.3,
          "out": 0.5,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://techcrunch.com/2025/04/09/elon-musks-ai-company-xai-launches-an-api-for-grok-3"
        }
      ],
      "note": "131K context; a low-cost reasoning-first model with adjustable 'Think' effort, launched alongside grok-3.",
      "last_checked": "2026-09-09"
    },
    "grok-3-mini-fast": {
      "name": "Grok 3 Mini Fast",
      "provider": "xai",
      "status": "legacy",
      "tier": "reasoning",
      "events": [
        {
          "date": "2025-04-09",
          "in": 0.6,
          "out": 4.0,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://techcrunch.com/2025/04/09/elon-musks-ai-company-xai-launches-an-api-for-grok-3"
        }
      ],
      "note": "Low-latency variant of grok-3-mini; despite the 'mini' name it is priced above grok-3's own output rate, reflecting a fast/high-throughput surcharge.",
      "last_checked": "2026-09-09"
    },
    "grok-4": {
      "name": "Grok 4",
      "provider": "xai",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2025-07-09",
          "in": 3.0,
          "out": 15.0,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://x.ai/news/grok-4"
        }
      ],
      "retired": "2026-05-15",
      "note": "256K context; native tool use; also sold as the pinned snapshot alias grok-4-0709 (the id xAI's own retirement notice uses). Price shown is the <128K-token tier; multiple third-party trackers describe a long-context tier"
    },
    "grok-code-fast-1": {
      "name": "Grok Code Fast 1",
      "provider": "xai",
      "status": "legacy",
      "tier": "other",
      "events": [
        {
          "date": "2025-08-28",
          "in": 0.2,
          "out": 1.5,
          "kind": "launch",
          "source": "https://x.ai/news/grok-code-fast-1"
        }
      ],
      "retired": "2026-05-15",
      "note": "Purpose-built agentic coding model (not a general chat model), 256K context per third-party trackers. Built on a new architecture trained on a code-heavy corpus; strongest on TypeScript, Python, Java, Rust, C++, Go."
    },
    "grok-4-fast-reasoning": {
      "name": "Grok 4 Fast (Reasoning)",
      "provider": "xai",
      "status": "legacy",
      "tier": "reasoning",
      "events": [
        {
          "date": "2025-09-19",
          "in": 0.2,
          "out": 0.5,
          "kind": "launch",
          "source": "https://x.ai/news/grok-4-fast"
        }
      ],
      "retired": "2026-05-15",
      "note": "2M-token context, the largest of any Grok text model to date. xAI claimed a 98% cost reduction versus grok-4 for equivalent frontier-benchmark performance via ~40% fewer 'thinking' tokens. Price doubles to $0.40/$1.00 pe"
    },
    "grok-4-fast-non-reasoning": {
      "name": "Grok 4 Fast (Non-Reasoning)",
      "provider": "xai",
      "status": "legacy",
      "tier": "mid",
      "events": [
        {
          "date": "2025-09-19",
          "in": 0.2,
          "out": 0.5,
          "kind": "launch",
          "source": "https://x.ai/news/grok-4-fast"
        }
      ],
      "retired": "2026-05-15",
      "note": "Same 2M context and pricing as grok-4-fast-reasoning, with reasoning disabled for lower-latency, non-chain-of-thought use."
    },
    "grok-4-1": {
      "name": "Grok 4.1",
      "provider": "xai",
      "status": "legacy",
      "tier": "flagship",
      "events": [
        {
          "date": "2025-11-17",
          "in": 3.0,
          "out": 15.0,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://llm-stats.com/models/grok-4.1-2025-11-17"
        }
      ],
      "retired": "2026-05-15",
      "note": "Conversational-quality refresh of grok-4 (lower hallucination rate, #1 on LMArena Text Arena at launch); announced publicly Nov 17, 2025 with API access opening Nov 19, 2025. No price change from grok-4."
    },
    "grok-4-1-fast-reasoning": {
      "name": "Grok 4.1 Fast (Reasoning)",
      "provider": "xai",
      "status": "legacy",
      "tier": "reasoning",
      "events": [
        {
          "date": "2025-11-19",
          "in": 0.2,
          "out": 0.5,
          "kind": "launch",
          "confidence": "secondary",
          "source": "https://venturebeat.com/ai/musks-xai-launches-grok-4-1-with-lower-hallucination-rate-on-the-web-and"
        }
      ],
      "retired": "2026-05-15",
      "note": "2M context; matches grok-4-fast pricing and tiering exactly; added the Agent Tools API (web search/code execution inside the reasoning loop). Launched in xAI's Enterprise API alongside a cut to agent-tool call pricing (m"
    },
    "grok-4-1-fast-non-reasoning": {
      "name": "Grok 4.1 Fast (Non-Reasoning)",
      "provider": "xai",
      "status": "legacy",
      "tier": "mid",
      "events": [
        {
          "date": "2025-11-19",
          "in": 0.2,
          "out": 0.5,
          "kind": "launch",
          "source": "https://docs.x.ai/developers/release-notes"
        }
      ],
      "retired": "2026-05-15",
      "note": "Same 2M context and pricing as grok-4-1-fast-reasoning, reasoning disabled."
    },
    "grok-4.20-0309-reasoning": {
      "name": "Grok 4.20 (Reasoning)",
      "provider": "xai",
      "status": "legacy",
      "tier": "reasoning",
      "events": [
        {
          "date": "2026-03-10",
          "in": 1.25,
          "out": 2.5,
          "kind": "launch",
          "source": "https://docs.x.ai/developers/release-notes"
        }
      ],
      "note": "1M context. Shipped as three simultaneous dated SKUs (reasoning / non-reasoning / multi-agent) succeeding grok-4.1 at under half its price ($1.25/$2.50 vs $3/$15) — briefly xAI's best general model until grok-4.3. Long-c",
      "last_checked": "2026-09-09"
    },
    "grok-4.20-0309-non-reasoning": {
      "name": "Grok 4.20 (Non-Reasoning)",
      "provider": "xai",
      "status": "legacy",
      "tier": "mid",
      "events": [
        {
          "date": "2026-03-10",
          "in": 1.25,
          "out": 2.5,
          "kind": "launch",
          "source": "https://docs.x.ai/developers/pricing"
        }
      ],
      "note": "Same 1M context and pricing as grok-4.20-0309-reasoning, reasoning disabled. This is the id xAI's May 15, 2026 retirement notice redirects grok-4-fast-non-reasoning and grok-4-1-fast-non-reasoning traffic to.",
      "last_checked": "2026-09-09"
    },
    "grok-build-0.1": {
      "name": "Grok Build 0.1",
      "provider": "xai",
      "status": "legacy",
      "tier": "other",
      "events": [
        {
          "date": "2026-05-19",
          "in": 1.0,
          "out": 2.0,
          "kind": "launch",
          "source": "https://docs.x.ai/developers/pricing"
        }
      ],
      "note": "256K context; agentic coding model built as a separate line from the main Grok chat family (not a fine-tune of grok-4.x). Became the retirement target for grok-code-fast-1 in xAI's current docs, and at $1/$2 is the cheap",
      "last_checked": "2026-09-09"
    }
  }
}