{"version":1,"updated":"2026-10-08","windowStart":"2025-07-01","scope":"Language models (chat, reasoning, completions, code, search, computer-use and deep-research IDs, and OpenAI fine-tune families) with a shutdown date on or after 1 July 2025, from four official pages: OpenAI's deprecations page, Anthropic's model deprecations doc, the Gemini API deprecations page and Vertex AI's model versions page. Audio, realtime, speech, transcription, image, video, embedding and moderation models are left out.","priceBasis":"US dollars per million tokens at the standard tier, text input, shortest context tier, as listed on the provider's pricing page on the date seen. A ratio is the replacement's price divided by the retiring model's; it ignores differences in how many tokens each model uses.","providers":{"openai":{"name":"OpenAI","platform":"OpenAI API","deprecations":"https://developers.openai.com/api/docs/deprecations","pricing":"https://developers.openai.com/api/docs/pricing","seen":"2026-10-08"},"anthropic":{"name":"Anthropic","platform":"Claude API","deprecations":"https://platform.claude.com/docs/en/about-claude/model-deprecations","pricing":"https://platform.claude.com/docs/en/about-claude/pricing","seen":"2026-10-08"},"gemini-api":{"name":"Google","platform":"Gemini API","deprecations":"https://ai.google.dev/gemini-api/docs/deprecations","pricing":"https://ai.google.dev/gemini-api/docs/pricing","seen":"2026-10-08"},"vertex-ai":{"name":"Google Cloud","platform":"Vertex AI (Gemini Enterprise Agent Platform)","deprecations":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/model-versions","pricing":"https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing","seen":"2026-10-08"}},"entries":[{"aliases":[],"kind":"model","announced":"2025-04-14","shutdownIsEarliest":false,"replacements":["gpt-4.1"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-4.5-preview","shutdown":"2025-07-14","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2025-01-21","shutdownIsEarliest":false,"replacements":["claude-opus-4-8"],"prices":[],"breaks":[{"short":"a non-default temperature, top_p or top_k returns a 400 error","text":"Claude 4.7 and later models return a 400 error when temperature, top_p or top_k is set to a non-default value. Anthropic's Python SDK v1.0 and later removes these parameters, so passing them raises a TypeError. Omit them and guide the model with the prompt.","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations"},{"short":"a newer tokenizer counts about 30% more tokens for the same text","text":"Claude 4.7 and later models use a newer tokenizer that produces approximately 30% more tokens for the same text, so the same prompt can cost more than the per-token ratio suggests. Measure on your own prompts.","source":"https://platform.claude.com/docs/en/about-claude/pricing"}],"notes":["Dates apply to the Claude API, Claude Platform on AWS and Microsoft Foundry. Amazon Bedrock and Google Cloud set their own retirement schedules."],"seen":"2026-10-08","provider":"anthropic","id":"claude-2.0","shutdown":"2025-07-21","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations","priceNote":"Not listed on Anthropic's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2025-01-21","shutdownIsEarliest":false,"replacements":["claude-opus-4-8"],"prices":[],"breaks":[{"short":"a non-default temperature, top_p or top_k returns a 400 error","text":"Claude 4.7 and later models return a 400 error when temperature, top_p or top_k is set to a non-default value. Anthropic's Python SDK v1.0 and later removes these parameters, so passing them raises a TypeError. Omit them and guide the model with the prompt.","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations"},{"short":"a newer tokenizer counts about 30% more tokens for the same text","text":"Claude 4.7 and later models use a newer tokenizer that produces approximately 30% more tokens for the same text, so the same prompt can cost more than the per-token ratio suggests. Measure on your own prompts.","source":"https://platform.claude.com/docs/en/about-claude/pricing"}],"notes":["Dates apply to the Claude API, Claude Platform on AWS and Microsoft Foundry. Amazon Bedrock and Google Cloud set their own retirement schedules."],"seen":"2026-10-08","provider":"anthropic","id":"claude-2.1","shutdown":"2025-07-21","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations","priceNote":"Not listed on Anthropic's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2025-01-21","shutdownIsEarliest":false,"replacements":["claude-sonnet-4-6"],"prices":[],"breaks":[],"notes":["Dates apply to the Claude API, Claude Platform on AWS and Microsoft Foundry. Amazon Bedrock and Google Cloud set their own retirement schedules."],"seen":"2026-10-08","provider":"anthropic","id":"claude-3-sonnet-20240229","shutdown":"2025-07-21","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations","priceNote":"Not listed on Anthropic's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2025-04-28","shutdownIsEarliest":false,"replacements":["o3"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"o1-preview","shutdown":"2025-07-28","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":false,"replacements":["gemini-2.5-flash-lite"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"vertex-ai","id":"gemini-1.5-flash-002","shutdown":"2025-09-24","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/model-versions","priceNote":"The retiring model is not listed on the Agent Platform pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":false,"replacements":["gemini-2.5-flash"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"vertex-ai","id":"gemini-1.5-pro-002","shutdown":"2025-09-24","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/model-versions","priceNote":"The retiring model is not listed on the Agent Platform pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2025-04-28","shutdownIsEarliest":false,"replacements":["o4-mini"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"o1-mini","shutdown":"2025-10-27","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2025-08-13","shutdownIsEarliest":false,"replacements":["claude-sonnet-4-6"],"prices":[],"breaks":[],"notes":["Dates apply to the Claude API, Claude Platform on AWS and Microsoft Foundry. Amazon Bedrock and Google Cloud set their own retirement schedules."],"seen":"2026-10-08","provider":"anthropic","id":"claude-3-5-sonnet-20240620","shutdown":"2025-10-28","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations","priceNote":"Not listed on Anthropic's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2025-08-13","shutdownIsEarliest":false,"replacements":["claude-sonnet-4-6"],"prices":[],"breaks":[],"notes":["Dates apply to the Claude API, Claude Platform on AWS and Microsoft Foundry. Amazon Bedrock and Google Cloud set their own retirement schedules."],"seen":"2026-10-08","provider":"anthropic","id":"claude-3-5-sonnet-20241022","shutdown":"2025-10-28","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations","priceNote":"Not listed on Anthropic's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":true,"replacements":["gemini-3.6-flash"],"prices":[],"breaks":[{"short":"a temperature below 1.0 may cause looping","text":"For Gemini 3 models Google recommends keeping temperature at its default of 1.0; setting it below 1.0 may cause looping or degraded performance.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"},{"short":"sending thinking_level and thinking_budget together returns a 400 error","text":"Gemini 3 uses thinking_level; sending thinking_level and the legacy thinking_budget in the same request returns a 400 error.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"}],"notes":["Google lists this as the earliest possible shutdown date and says it will give the exact date with advance notice."],"seen":"2026-10-08","provider":"gemini-api","id":"gemini-2.5-flash-preview-05-20","shutdown":"2025-11-18","source":"https://ai.google.dev/gemini-api/docs/deprecations","priceNote":"The retiring model is not listed on the Gemini API pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":true,"replacements":["gemini-3.1-pro-preview"],"prices":[],"breaks":[{"short":"a temperature below 1.0 may cause looping","text":"For Gemini 3 models Google recommends keeping temperature at its default of 1.0; setting it below 1.0 may cause looping or degraded performance.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"},{"short":"sending thinking_level and thinking_budget together returns a 400 error","text":"Gemini 3 uses thinking_level; sending thinking_level and the legacy thinking_budget in the same request returns a 400 error.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"}],"notes":["Google lists this as the earliest possible shutdown date and says it will give the exact date with advance notice."],"seen":"2026-10-08","provider":"gemini-api","id":"gemini-2.5-pro-preview-03-25","shutdown":"2025-12-02","source":"https://ai.google.dev/gemini-api/docs/deprecations","priceNote":"The retiring model is not listed on the Gemini API pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":true,"replacements":["gemini-3.1-pro-preview"],"prices":[],"breaks":[{"short":"a temperature below 1.0 may cause looping","text":"For Gemini 3 models Google recommends keeping temperature at its default of 1.0; setting it below 1.0 may cause looping or degraded performance.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"},{"short":"sending thinking_level and thinking_budget together returns a 400 error","text":"Gemini 3 uses thinking_level; sending thinking_level and the legacy thinking_budget in the same request returns a 400 error.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"}],"notes":["Google lists this as the earliest possible shutdown date and says it will give the exact date with advance notice."],"seen":"2026-10-08","provider":"gemini-api","id":"gemini-2.5-pro-preview-05-06","shutdown":"2025-12-02","source":"https://ai.google.dev/gemini-api/docs/deprecations","priceNote":"The retiring model is not listed on the Gemini API pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":true,"replacements":["gemini-3.1-pro-preview"],"prices":[],"breaks":[{"short":"a temperature below 1.0 may cause looping","text":"For Gemini 3 models Google recommends keeping temperature at its default of 1.0; setting it below 1.0 may cause looping or degraded performance.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"},{"short":"sending thinking_level and thinking_budget together returns a 400 error","text":"Gemini 3 uses thinking_level; sending thinking_level and the legacy thinking_budget in the same request returns a 400 error.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"}],"notes":["Google lists this as the earliest possible shutdown date and says it will give the exact date with advance notice."],"seen":"2026-10-08","provider":"gemini-api","id":"gemini-2.5-pro-preview-06-05","shutdown":"2025-12-02","source":"https://ai.google.dev/gemini-api/docs/deprecations","priceNote":"The retiring model is not listed on the Gemini API pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":true,"replacements":["gemini-2.5-flash-lite"],"prices":[],"breaks":[],"notes":["Google lists this as the earliest possible shutdown date and says it will give the exact date with advance notice."],"seen":"2026-10-08","provider":"gemini-api","id":"gemini-2.0-flash-lite-preview","shutdown":"2025-12-09","source":"https://ai.google.dev/gemini-api/docs/deprecations","priceNote":"The retiring model is not listed on the Gemini API pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":true,"replacements":["gemini-2.5-flash-lite"],"prices":[],"breaks":[],"notes":["Google lists this as the earliest possible shutdown date and says it will give the exact date with advance notice."],"seen":"2026-10-08","provider":"gemini-api","id":"gemini-2.0-flash-lite-preview-02-05","shutdown":"2025-12-09","source":"https://ai.google.dev/gemini-api/docs/deprecations","priceNote":"The retiring model is not listed on the Gemini API pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2025-06-30","shutdownIsEarliest":false,"replacements":["claude-opus-4-8"],"prices":[],"breaks":[{"short":"a non-default temperature, top_p or top_k returns a 400 error","text":"Claude 4.7 and later models return a 400 error when temperature, top_p or top_k is set to a non-default value. Anthropic's Python SDK v1.0 and later removes these parameters, so passing them raises a TypeError. Omit them and guide the model with the prompt.","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations"},{"short":"a newer tokenizer counts about 30% more tokens for the same text","text":"Claude 4.7 and later models use a newer tokenizer that produces approximately 30% more tokens for the same text, so the same prompt can cost more than the per-token ratio suggests. Measure on your own prompts.","source":"https://platform.claude.com/docs/en/about-claude/pricing"}],"notes":["Dates apply to the Claude API, Claude Platform on AWS and Microsoft Foundry. Amazon Bedrock and Google Cloud set their own retirement schedules."],"seen":"2026-10-08","provider":"anthropic","id":"claude-3-opus-20240229","shutdown":"2026-01-05","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations","priceNote":"Not listed on Anthropic's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2025-11-17","shutdownIsEarliest":false,"replacements":["gpt-5-codex-mini"],"prices":[],"breaks":[{"short":"the legacy local shell tool is no longer supported","text":"The legacy local shell tool, available only with codex-mini-latest, is no longer supported; OpenAI points to its latest shell tool.","source":"https://developers.openai.com/api/docs/deprecations"}],"notes":[],"seen":"2026-10-08","provider":"openai","id":"codex-mini-latest","shutdown":"2026-02-12","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":true,"replacements":["gemini-3.6-flash"],"prices":[],"breaks":[{"short":"a temperature below 1.0 may cause looping","text":"For Gemini 3 models Google recommends keeping temperature at its default of 1.0; setting it below 1.0 may cause looping or degraded performance.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"},{"short":"sending thinking_level and thinking_budget together returns a 400 error","text":"Gemini 3 uses thinking_level; sending thinking_level and the legacy thinking_budget in the same request returns a 400 error.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"}],"notes":["Google lists this as the earliest possible shutdown date and says it will give the exact date with advance notice."],"seen":"2026-10-08","provider":"gemini-api","id":"gemini-2.5-flash-preview-09-25","shutdown":"2026-02-17","source":"https://ai.google.dev/gemini-api/docs/deprecations","priceNote":"The retiring model is not listed on the Gemini API pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2025-11-18","shutdownIsEarliest":false,"replacements":["gpt-5.1-chat-latest"],"prices":[],"breaks":[],"notes":["The replacement OpenAI named, gpt-5.1-chat-latest, was itself shut down on 23 July 2026."],"seen":"2026-10-08","provider":"openai","id":"chatgpt-4o-latest","shutdown":"2026-02-17","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2025-12-19","shutdownIsEarliest":false,"replacements":["claude-haiku-4-5-20251001"],"prices":[{"replacement":"claude-haiku-4-5-20251001","retiring":{"input":0.8,"output":4},"replacementPrice":{"input":1,"output":5},"ratio":{"input":1.25,"output":1.25}}],"breaks":[],"notes":["Dates apply to the Claude API, Claude Platform on AWS and Microsoft Foundry. Amazon Bedrock and Google Cloud set their own retirement schedules."],"seen":"2026-10-08","provider":"anthropic","id":"claude-3-5-haiku-20241022","shutdown":"2026-02-19","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations","pricingSource":"https://platform.claude.com/docs/en/about-claude/pricing","pricingSeen":"2026-10-08"},{"aliases":[],"kind":"model","announced":"2025-10-28","shutdownIsEarliest":false,"replacements":["claude-sonnet-4-6"],"prices":[],"breaks":[],"notes":["Dates apply to the Claude API, Claude Platform on AWS and Microsoft Foundry. Amazon Bedrock and Google Cloud set their own retirement schedules."],"seen":"2026-10-08","provider":"anthropic","id":"claude-3-7-sonnet-20250219","shutdown":"2026-02-19","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations","priceNote":"Not listed on Anthropic's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":true,"replacements":["gemini-3.1-pro-preview"],"prices":[],"breaks":[],"notes":["Google lists this as the earliest possible shutdown date and says it will give the exact date with advance notice."],"seen":"2026-10-08","provider":"gemini-api","id":"gemini-3-pro-preview","shutdown":"2026-03-09","source":"https://ai.google.dev/gemini-api/docs/deprecations","priceNote":"The retiring model is not listed on the Gemini API pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":["gpt-4-turbo-preview","gpt-4-turbo-preview-completions"],"kind":"model","announced":"2025-09-26","shutdownIsEarliest":false,"replacements":["gpt-5","gpt-4.1"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-4-0125-preview","shutdown":"2026-03-26","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2025-09-26","shutdownIsEarliest":false,"replacements":["gpt-5","gpt-4.1"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-4-0314","shutdown":"2026-03-26","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":true,"replacements":["gemini-3.1-flash-lite"],"prices":[],"breaks":[{"short":"a temperature below 1.0 may cause looping","text":"For Gemini 3 models Google recommends keeping temperature at its default of 1.0; setting it below 1.0 may cause looping or degraded performance.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"},{"short":"sending thinking_level and thinking_budget together returns a 400 error","text":"Gemini 3 uses thinking_level; sending thinking_level and the legacy thinking_budget in the same request returns a 400 error.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"}],"notes":["Google lists this as the earliest possible shutdown date and says it will give the exact date with advance notice."],"seen":"2026-10-08","provider":"gemini-api","id":"gemini-2.5-flash-lite-preview-09-2025","shutdown":"2026-03-31","source":"https://ai.google.dev/gemini-api/docs/deprecations","priceNote":"The retiring model is not listed on the Gemini API pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2026-02-19","shutdownIsEarliest":false,"replacements":["claude-haiku-4-5-20251001"],"prices":[],"breaks":[],"notes":["Dates apply to the Claude API, Claude Platform on AWS and Microsoft Foundry. Amazon Bedrock and Google Cloud set their own retirement schedules."],"seen":"2026-10-08","provider":"anthropic","id":"claude-3-haiku-20240307","shutdown":"2026-04-20","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations","priceNote":"Not listed on Anthropic's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":true,"replacements":["gemini-3.1-flash-lite"],"prices":[],"breaks":[],"notes":["Google lists this as the earliest possible shutdown date and says it will give the exact date with advance notice."],"seen":"2026-10-08","provider":"gemini-api","id":"gemini-3.1-flash-lite-preview","shutdown":"2026-05-25","source":"https://ai.google.dev/gemini-api/docs/deprecations","priceNote":"The retiring model is not listed on the Gemini API pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":true,"replacements":["gemini-3.6-flash"],"prices":[],"breaks":[{"short":"a temperature below 1.0 may cause looping","text":"For Gemini 3 models Google recommends keeping temperature at its default of 1.0; setting it below 1.0 may cause looping or degraded performance.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"},{"short":"sending thinking_level and thinking_budget together returns a 400 error","text":"Gemini 3 uses thinking_level; sending thinking_level and the legacy thinking_budget in the same request returns a 400 error.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"}],"notes":["Google lists this as the earliest possible shutdown date and says it will give the exact date with advance notice."],"seen":"2026-10-08","provider":"gemini-api","id":"gemini-2.0-flash","shutdown":"2026-06-01","source":"https://ai.google.dev/gemini-api/docs/deprecations","priceNote":"The retiring model is not listed on the Gemini API pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":true,"replacements":["gemini-3.6-flash"],"prices":[],"breaks":[{"short":"a temperature below 1.0 may cause looping","text":"For Gemini 3 models Google recommends keeping temperature at its default of 1.0; setting it below 1.0 may cause looping or degraded performance.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"},{"short":"sending thinking_level and thinking_budget together returns a 400 error","text":"Gemini 3 uses thinking_level; sending thinking_level and the legacy thinking_budget in the same request returns a 400 error.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"}],"notes":["Google lists this as the earliest possible shutdown date and says it will give the exact date with advance notice."],"seen":"2026-10-08","provider":"gemini-api","id":"gemini-2.0-flash-001","shutdown":"2026-06-01","source":"https://ai.google.dev/gemini-api/docs/deprecations","priceNote":"The retiring model is not listed on the Gemini API pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":true,"replacements":["gemini-3.1-flash-lite"],"prices":[],"breaks":[{"short":"a temperature below 1.0 may cause looping","text":"For Gemini 3 models Google recommends keeping temperature at its default of 1.0; setting it below 1.0 may cause looping or degraded performance.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"},{"short":"sending thinking_level and thinking_budget together returns a 400 error","text":"Gemini 3 uses thinking_level; sending thinking_level and the legacy thinking_budget in the same request returns a 400 error.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"}],"notes":["Google lists this as the earliest possible shutdown date and says it will give the exact date with advance notice."],"seen":"2026-10-08","provider":"gemini-api","id":"gemini-2.0-flash-lite","shutdown":"2026-06-01","source":"https://ai.google.dev/gemini-api/docs/deprecations","priceNote":"The retiring model is not listed on the Gemini API pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":true,"replacements":["gemini-3.1-flash-lite"],"prices":[],"breaks":[{"short":"a temperature below 1.0 may cause looping","text":"For Gemini 3 models Google recommends keeping temperature at its default of 1.0; setting it below 1.0 may cause looping or degraded performance.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"},{"short":"sending thinking_level and thinking_budget together returns a 400 error","text":"Gemini 3 uses thinking_level; sending thinking_level and the legacy thinking_budget in the same request returns a 400 error.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"}],"notes":["Google lists this as the earliest possible shutdown date and says it will give the exact date with advance notice."],"seen":"2026-10-08","provider":"gemini-api","id":"gemini-2.0-flash-lite-001","shutdown":"2026-06-01","source":"https://ai.google.dev/gemini-api/docs/deprecations","priceNote":"The retiring model is not listed on the Gemini API pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":false,"replacements":["gemini-3.1-flash-lite"],"prices":[],"breaks":[{"short":"temperature, top_p and top_k are deprecated and should be removed","text":"For Gemini 3.x models, temperature, top_p and top_k are deprecated: Google says to remove them from every request and to use a system instruction for determinism.","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/migrate"},{"short":"thinking_budget gives way to thinking_level","text":"thinking_budget is deprecated on Gemini 3.x in favour of thinking_level.","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/migrate"},{"short":"a function response that does not match its call returns an empty answer, not an error","text":"Every FunctionResponse must carry the id and name of its FunctionCall, exactly one per call. Google says the API does not yet return an error for a mismatch; the model returns an empty response with finish_reason STOP instead.","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/migrate"}],"notes":[],"seen":"2026-10-08","provider":"vertex-ai","id":"gemini-2.0-flash","shutdown":"2026-06-01","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/model-versions","priceNote":"The retiring model is not listed on the Agent Platform pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":false,"replacements":["gemini-3.1-flash-lite"],"prices":[],"breaks":[{"short":"temperature, top_p and top_k are deprecated and should be removed","text":"For Gemini 3.x models, temperature, top_p and top_k are deprecated: Google says to remove them from every request and to use a system instruction for determinism.","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/migrate"},{"short":"thinking_budget gives way to thinking_level","text":"thinking_budget is deprecated on Gemini 3.x in favour of thinking_level.","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/migrate"},{"short":"a function response that does not match its call returns an empty answer, not an error","text":"Every FunctionResponse must carry the id and name of its FunctionCall, exactly one per call. Google says the API does not yet return an error for a mismatch; the model returns an empty response with finish_reason STOP instead.","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/migrate"}],"notes":[],"seen":"2026-10-08","provider":"vertex-ai","id":"gemini-2.0-flash-lite","shutdown":"2026-06-01","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/model-versions","priceNote":"The retiring model is not listed on the Agent Platform pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2026-04-14","shutdownIsEarliest":false,"replacements":["claude-opus-4-8"],"prices":[{"replacement":"claude-opus-4-8","retiring":{"input":15,"output":75},"replacementPrice":{"input":5,"output":25},"ratio":{"input":0.333,"output":0.333}}],"breaks":[{"short":"a non-default temperature, top_p or top_k returns a 400 error","text":"Claude 4.7 and later models return a 400 error when temperature, top_p or top_k is set to a non-default value. Anthropic's Python SDK v1.0 and later removes these parameters, so passing them raises a TypeError. Omit them and guide the model with the prompt.","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations"},{"short":"a newer tokenizer counts about 30% more tokens for the same text","text":"Claude 4.7 and later models use a newer tokenizer that produces approximately 30% more tokens for the same text, so the same prompt can cost more than the per-token ratio suggests. Measure on your own prompts.","source":"https://platform.claude.com/docs/en/about-claude/pricing"}],"notes":["Dates apply to the Claude API, Claude Platform on AWS and Microsoft Foundry. Amazon Bedrock and Google Cloud set their own retirement schedules."],"seen":"2026-10-08","provider":"anthropic","id":"claude-opus-4-20250514","shutdown":"2026-06-15","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations","pricingSource":"https://platform.claude.com/docs/en/about-claude/pricing","pricingSeen":"2026-10-08"},{"aliases":[],"kind":"model","announced":"2026-04-14","shutdownIsEarliest":false,"replacements":["claude-sonnet-4-6"],"prices":[{"replacement":"claude-sonnet-4-6","retiring":{"input":3,"output":15},"replacementPrice":{"input":3,"output":15},"ratio":{"input":1,"output":1}}],"breaks":[],"notes":["Dates apply to the Claude API, Claude Platform on AWS and Microsoft Foundry. Amazon Bedrock and Google Cloud set their own retirement schedules."],"seen":"2026-10-08","provider":"anthropic","id":"claude-sonnet-4-20250514","shutdown":"2026-06-15","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations","pricingSource":"https://platform.claude.com/docs/en/about-claude/pricing","pricingSeen":"2026-10-08"},{"aliases":["computer-use-preview"],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-terra"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"computer-use-preview-2025-03-11","shutdown":"2026-07-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-terra"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-4o-mini-search-preview-2025-03-11","shutdown":"2026-07-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-terra"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-4o-search-preview-2025-03-11","shutdown":"2026-07-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-5-chat-latest","shutdown":"2026-07-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-5-codex","shutdown":"2026-07-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-5.1-chat-latest","shutdown":"2026-07-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-5.1-codex","shutdown":"2026-07-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-5.1-codex-max","shutdown":"2026-07-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-terra"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-5.1-codex-mini","shutdown":"2026-07-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-5.2-codex","shutdown":"2026-07-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":["o3-deep-research"],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"o3-deep-research-2025-06-26","shutdown":"2026-07-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":["o4-mini-deep-research"],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"o4-mini-deep-research-2025-06-26","shutdown":"2026-07-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":true,"replacements":["gemini-3.8-flash"],"prices":[],"breaks":[{"short":"a temperature below 1.0 may cause looping","text":"For Gemini 3 models Google recommends keeping temperature at its default of 1.0; setting it below 1.0 may cause looping or degraded performance.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"},{"short":"sending thinking_level and thinking_budget together returns a 400 error","text":"Gemini 3 uses thinking_level; sending thinking_level and the legacy thinking_budget in the same request returns a 400 error.","source":"https://ai.google.dev/gemini-api/docs/gemini-3"}],"notes":["Google lists this as the earliest possible shutdown date and says it will give the exact date with advance notice."],"seen":"2026-10-08","provider":"gemini-api","id":"gemini-2.5-computer-use-preview-10-2025","shutdown":"2026-07-28","source":"https://ai.google.dev/gemini-api/docs/deprecations","priceNote":"The retiring model is not listed on the Gemini API pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":["claude-opus-4-1"],"kind":"model","announced":"2026-06-05","shutdownIsEarliest":false,"replacements":["claude-opus-4-8"],"prices":[{"replacement":"claude-opus-4-8","retiring":{"input":15,"output":75},"replacementPrice":{"input":5,"output":25},"ratio":{"input":0.333,"output":0.333}}],"breaks":[{"short":"a non-default temperature, top_p or top_k returns a 400 error","text":"Claude 4.7 and later models return a 400 error when temperature, top_p or top_k is set to a non-default value. Anthropic's Python SDK v1.0 and later removes these parameters, so passing them raises a TypeError. Omit them and guide the model with the prompt.","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations"},{"short":"a newer tokenizer counts about 30% more tokens for the same text","text":"Claude 4.7 and later models use a newer tokenizer that produces approximately 30% more tokens for the same text, so the same prompt can cost more than the per-token ratio suggests. Measure on your own prompts.","source":"https://platform.claude.com/docs/en/about-claude/pricing"}],"notes":["Dates apply to the Claude API, Claude Platform on AWS and Microsoft Foundry. Amazon Bedrock and Google Cloud set their own retirement schedules."],"seen":"2026-10-08","provider":"anthropic","id":"claude-opus-4-1-20250805","shutdown":"2026-08-05","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations","pricingSource":"https://platform.claude.com/docs/en/about-claude/pricing","pricingSeen":"2026-10-08","aliasSource":"https://platform.claude.com/docs/en/about-claude/models/model-ids-and-versions"},{"aliases":[],"kind":"model","announced":"2026-05-08","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-5.2-chat-latest","shutdown":"2026-08-10","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2026-05-08","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-5.3-chat-latest","shutdown":"2026-08-10","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":[],"kind":"model","announced":"2025-09-26","shutdownIsEarliest":false,"replacements":["gpt-5.6-terra"],"prices":[{"replacement":"gpt-5.6-terra","retiring":{"input":0.4,"output":0.4},"replacementPrice":{"input":2,"output":12},"ratio":{"input":5,"output":30}}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"babbage-002","shutdown":"2026-09-28","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08"},{"aliases":[],"kind":"model","announced":"2025-09-26","shutdownIsEarliest":false,"replacements":["gpt-5.6-terra"],"prices":[{"replacement":"gpt-5.6-terra","retiring":{"input":2,"output":2},"replacementPrice":{"input":2,"output":12},"ratio":{"input":1,"output":6}}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"davinci-002","shutdown":"2026-09-28","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08"},{"aliases":[],"kind":"model","announced":"2025-09-26","shutdownIsEarliest":false,"replacements":["gpt-5.6-terra"],"prices":[{"replacement":"gpt-5.6-terra","retiring":{"input":1,"output":2},"replacementPrice":{"input":2,"output":12},"ratio":{"input":2,"output":6}}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-3.5-turbo-1106","shutdown":"2026-09-28","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08"},{"aliases":[],"kind":"model","announced":"2025-09-26","shutdownIsEarliest":false,"replacements":["gpt-5.6-terra"],"prices":[{"replacement":"gpt-5.6-terra","retiring":{"input":1.5,"output":2},"replacementPrice":{"input":2,"output":12},"ratio":{"input":1.33,"output":6}}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-3.5-turbo-instruct","shutdown":"2026-09-28","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08"},{"aliases":[],"kind":"model","announced":"2026-09-11","shutdownIsEarliest":false,"replacements":[],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-5.4-cyber","shutdown":"2026-10-01","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"OpenAI names no specific replacement model, so no ratio is given.","replacementNote":"The most capable cyber model available to you."},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":false,"replacements":["gemini-3.8-flash","gemini-3.5-flash-lite","gemini-3.1-flash-lite"],"prices":[{"replacement":"gemini-3.8-flash","retiring":{"input":0.3,"output":2.5},"replacementPrice":{"input":0.75,"output":3.75},"ratio":{"input":2.5,"output":1.5},"until":"2026-12-31"},{"replacement":"gemini-3.8-flash","retiring":{"input":0.3,"output":2.5},"replacementPrice":{"input":1.5,"output":7.5},"ratio":{"input":5,"output":3},"from":"2027-01-01"},{"replacement":"gemini-3.5-flash-lite","retiring":{"input":0.3,"output":2.5},"replacementPrice":{"input":0.3,"output":2.5},"ratio":{"input":1,"output":1}},{"replacement":"gemini-3.1-flash-lite","retiring":{"input":0.3,"output":2.5},"replacementPrice":{"input":0.25,"output":1.5},"ratio":{"input":0.833,"output":0.6}}],"breaks":[{"short":"temperature, top_p and top_k are deprecated and should be removed","text":"For Gemini 3.x models, temperature, top_p and top_k are deprecated: Google says to remove them from every request and to use a system instruction for determinism.","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/migrate"},{"short":"thinking_budget gives way to thinking_level","text":"thinking_budget is deprecated on Gemini 3.x in favour of thinking_level.","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/migrate"},{"short":"a function response that does not match its call returns an empty answer, not an error","text":"Every FunctionResponse must carry the id and name of its FunctionCall, exactly one per call. Google says the API does not yet return an error for a mismatch; the model returns an empty response with finish_reason STOP instead.","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/migrate"}],"notes":["On the Gemini API (ai.google.dev) the same model ID had no shutdown date announced when checked on 8 October 2026; Google limits 2.5 access there to projects that used it before. This date is for Vertex AI, now Gemini Enterprise Agent Platform.","Google says retirement dates on this page may be extended but will not be moved earlier."],"seen":"2026-10-08","provider":"vertex-ai","id":"gemini-2.5-flash","shutdown":"2026-10-20","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/model-versions","pricingSource":"https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing","pricingSeen":"2026-10-08"},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":false,"replacements":["gemini-3.8-flash","gemini-3.1-flash-lite","Gemma 4"],"prices":[{"replacement":"gemini-3.8-flash","retiring":{"input":0.1,"output":0.4},"replacementPrice":{"input":0.75,"output":3.75},"ratio":{"input":7.5,"output":9.38},"until":"2026-12-31"},{"replacement":"gemini-3.8-flash","retiring":{"input":0.1,"output":0.4},"replacementPrice":{"input":1.5,"output":7.5},"ratio":{"input":15,"output":18.8},"from":"2027-01-01"},{"replacement":"gemini-3.1-flash-lite","retiring":{"input":0.1,"output":0.4},"replacementPrice":{"input":0.25,"output":1.5},"ratio":{"input":2.5,"output":3.75}}],"breaks":[{"short":"temperature, top_p and top_k are deprecated and should be removed","text":"For Gemini 3.x models, temperature, top_p and top_k are deprecated: Google says to remove them from every request and to use a system instruction for determinism.","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/migrate"},{"short":"thinking_budget gives way to thinking_level","text":"thinking_budget is deprecated on Gemini 3.x in favour of thinking_level.","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/migrate"},{"short":"a function response that does not match its call returns an empty answer, not an error","text":"Every FunctionResponse must carry the id and name of its FunctionCall, exactly one per call. Google says the API does not yet return an error for a mismatch; the model returns an empty response with finish_reason STOP instead.","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/migrate"}],"notes":["On the Gemini API (ai.google.dev) the same model ID had no shutdown date announced when checked on 8 October 2026; Google limits 2.5 access there to projects that used it before. This date is for Vertex AI, now Gemini Enterprise Agent Platform.","Google says retirement dates on this page may be extended but will not be moved earlier."],"seen":"2026-10-08","provider":"vertex-ai","id":"gemini-2.5-flash-lite","shutdown":"2026-10-20","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/model-versions","pricingSource":"https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing","pricingSeen":"2026-10-08"},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":false,"replacements":["gemini-3.8-flash","gemini-3.5-flash"],"prices":[{"replacement":"gemini-3.8-flash","retiring":{"input":1.25,"output":10},"replacementPrice":{"input":0.75,"output":3.75},"ratio":{"input":0.6,"output":0.375},"until":"2026-12-31"},{"replacement":"gemini-3.8-flash","retiring":{"input":1.25,"output":10},"replacementPrice":{"input":1.5,"output":7.5},"ratio":{"input":1.2,"output":0.75},"from":"2027-01-01"},{"replacement":"gemini-3.5-flash","retiring":{"input":1.25,"output":10},"replacementPrice":{"input":1.5,"output":9},"ratio":{"input":1.2,"output":0.9}}],"breaks":[{"short":"temperature, top_p and top_k are deprecated and should be removed","text":"For Gemini 3.x models, temperature, top_p and top_k are deprecated: Google says to remove them from every request and to use a system instruction for determinism.","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/migrate"},{"short":"thinking_budget gives way to thinking_level","text":"thinking_budget is deprecated on Gemini 3.x in favour of thinking_level.","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/migrate"},{"short":"a function response that does not match its call returns an empty answer, not an error","text":"Every FunctionResponse must carry the id and name of its FunctionCall, exactly one per call. Google says the API does not yet return an error for a mismatch; the model returns an empty response with finish_reason STOP instead.","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/migrate"}],"notes":["On the Gemini API (ai.google.dev) the same model ID had no shutdown date announced when checked on 8 October 2026; Google limits 2.5 access there to projects that used it before. This date is for Vertex AI, now Gemini Enterprise Agent Platform.","Google says retirement dates on this page may be extended but will not be moved earlier."],"seen":"2026-10-08","provider":"vertex-ai","id":"gemini-2.5-pro","shutdown":"2026-10-20","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/model-versions","pricingSource":"https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing","pricingSeen":"2026-10-08"},{"aliases":[],"kind":"fine-tune","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-terra"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"ft-babbage-002","shutdown":"2026-10-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"A fine-tune's replacement is a base model you would fine-tune again or prompt instead, so a per-token ratio would compare unlike things; none is given.","finetunePrefixes":["ft:babbage-002:"]},{"aliases":[],"kind":"fine-tune","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-terra"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"ft-davinci-002","shutdown":"2026-10-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"A fine-tune's replacement is a base model you would fine-tune again or prompt instead, so a per-token ratio would compare unlike things; none is given.","finetunePrefixes":["ft:davinci-002:"]},{"aliases":[],"kind":"fine-tune","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-terra"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"ft-gpt-3.5-turbo","shutdown":"2026-10-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"A fine-tune's replacement is a base model you would fine-tune again or prompt instead, so a per-token ratio would compare unlike things; none is given.","finetunePrefixes":["ft:gpt-3.5-turbo:","ft:gpt-3.5-turbo-"]},{"aliases":[],"kind":"fine-tune","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"ft-gpt-4","shutdown":"2026-10-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"A fine-tune's replacement is a base model you would fine-tune again or prompt instead, so a per-token ratio would compare unlike things; none is given.","finetunePrefixes":["ft:gpt-4:","ft:gpt-4-0613:"]},{"aliases":[],"kind":"fine-tune","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-luna"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"ft-gpt-4.1-nano-2025-04-14","shutdown":"2026-10-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"A fine-tune's replacement is a base model you would fine-tune again or prompt instead, so a per-token ratio would compare unlike things; none is given.","finetunePrefixes":["ft:gpt-4.1-nano-2025-04-14:"]},{"aliases":[],"kind":"fine-tune","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-terra"],"prices":[],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"ft-o4-mini-2025-04-16","shutdown":"2026-10-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"A fine-tune's replacement is a base model you would fine-tune again or prompt instead, so a per-token ratio would compare unlike things; none is given.","finetunePrefixes":["ft:o4-mini-2025-04-16:"]},{"aliases":["gpt-3.5-turbo","gpt-3.5-turbo-completions"],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-terra"],"prices":[{"replacement":"gpt-5.6-terra","retiring":{"input":0.5,"output":1.5},"replacementPrice":{"input":2,"output":12},"ratio":{"input":4,"output":8}}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-3.5-turbo-0125","shutdown":"2026-10-23","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08"},{"aliases":["gpt-4","gpt-4-0613-completions","gpt-4-completions"],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[{"replacement":"gpt-5.6-sol","retiring":{"input":30,"output":60},"replacementPrice":{"input":4,"output":20},"ratio":{"input":0.133,"output":0.333},"note":"OpenAI lists gpt-5.6-sol's $4 / $20 as promotional pricing, available at least through 21 November 2026 (its model page, read 8 October 2026)."}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-4-0613","shutdown":"2026-10-23","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08"},{"aliases":[],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[],"breaks":[],"notes":["OpenAI's page also lists gpt-4-1106-preview under the shutdown of 26 March 2026 (announced 26 September 2025); this page uses the later announcement of 22 April 2026, which gives 23 October 2026."],"seen":"2026-10-08","provider":"openai","id":"gpt-4-1106-preview","shutdown":"2026-10-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"Not listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given."},{"aliases":["gpt-4-turbo-2024-04-09","gpt-4-turbo-completions"],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[{"replacement":"gpt-5.6-sol","retiring":{"input":10,"output":30},"replacementPrice":{"input":4,"output":20},"ratio":{"input":0.4,"output":0.667},"note":"OpenAI lists gpt-5.6-sol's $4 / $20 as promotional pricing, available at least through 21 November 2026 (its model page, read 8 October 2026)."}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-4-turbo","shutdown":"2026-10-23","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08"},{"aliases":["gpt-4.1-nano-2025-04-14"],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-luna"],"prices":[{"replacement":"gpt-5.6-luna","retiring":{"input":0.1,"output":0.4},"replacementPrice":{"input":0.2,"output":1.2},"ratio":{"input":2,"output":3}}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-4.1-nano","shutdown":"2026-10-23","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08"},{"aliases":[],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[{"replacement":"gpt-5.6-sol","retiring":{"input":5,"output":15},"replacementPrice":{"input":4,"output":20},"ratio":{"input":0.8,"output":1.33},"note":"OpenAI lists gpt-5.6-sol's $4 / $20 as promotional pricing, available at least through 21 November 2026 (its model page, read 8 October 2026)."}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-4o-2024-05-13","shutdown":"2026-10-23","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08"},{"aliases":["o1"],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[{"replacement":"gpt-5.6-sol","retiring":{"input":15,"output":60},"replacementPrice":{"input":4,"output":20},"ratio":{"input":0.267,"output":0.333},"note":"OpenAI lists gpt-5.6-sol's $4 / $20 as promotional pricing, available at least through 21 November 2026 (its model page, read 8 October 2026)."}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"o1-2024-12-17","shutdown":"2026-10-23","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08"},{"aliases":["o1-pro"],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[],"breaks":[{"short":"the replacement needs reasoning.mode set to pro","text":"OpenAI's replacement is gpt-5.6-sol with reasoning.mode set to pro, not the default mode.","source":"https://developers.openai.com/api/docs/deprecations"}],"notes":[],"seen":"2026-10-08","provider":"openai","id":"o1-pro-2025-03-19","shutdown":"2026-10-23","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"No price for gpt-5.6-sol with reasoning.mode pro was listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given.","replacementNote":"reasoning.mode: pro"},{"aliases":["o3-mini"],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[{"replacement":"gpt-5.6-sol","retiring":{"input":1.1,"output":4.4},"replacementPrice":{"input":4,"output":20},"ratio":{"input":3.64,"output":4.55},"note":"OpenAI lists gpt-5.6-sol's $4 / $20 as promotional pricing, available at least through 21 November 2026 (its model page, read 8 October 2026)."}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"o3-mini-2025-01-31","shutdown":"2026-10-23","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08"},{"aliases":["o4-mini"],"kind":"model","announced":"2026-04-22","shutdownIsEarliest":false,"replacements":["gpt-5.6-terra"],"prices":[{"replacement":"gpt-5.6-terra","retiring":{"input":1.1,"output":4.4},"replacementPrice":{"input":2,"output":12},"ratio":{"input":1.82,"output":2.73}}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"o4-mini-2025-04-16","shutdown":"2026-10-23","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08"},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":false,"replacements":["gemini-3.8-flash"],"prices":[{"replacement":"gemini-3.8-flash","retiring":{"input":0.75,"output":3.75},"replacementPrice":{"input":0.75,"output":3.75},"ratio":{"input":1,"output":1},"note":"Agent Platform lists both models at $0.75 / $3.75 through 31 December 2026 and $1.50 / $7.50 from 1 January 2027."}],"breaks":[],"notes":["A short-term availability model: Google posts a fixed retirement date at least 45 days ahead.","On the Gemini API (ai.google.dev) the same model ID had no shutdown date announced when checked on 8 October 2026."],"seen":"2026-10-08","provider":"vertex-ai","id":"gemini-3.6-flash","shutdown":"2026-11-19","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/model-versions","pricingSource":"https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing","pricingSeen":"2026-10-08"},{"aliases":["claude-sonnet-4-5"],"kind":"model","announced":"2026-09-30","shutdownIsEarliest":false,"replacements":["claude-sonnet-5-5"],"prices":[{"replacement":"claude-sonnet-5-5","retiring":{"input":3,"output":15},"replacementPrice":{"input":2,"output":10},"ratio":{"input":0.667,"output":0.667}}],"breaks":[{"short":"a non-default temperature, top_p or top_k returns a 400 error","text":"Claude 4.7 and later models return a 400 error when temperature, top_p or top_k is set to a non-default value. Anthropic's Python SDK v1.0 and later removes these parameters, so passing them raises a TypeError. Omit them and guide the model with the prompt.","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations"},{"short":"a newer tokenizer counts about 30% more tokens for the same text","text":"Claude 4.7 and later models use a newer tokenizer that produces approximately 30% more tokens for the same text, so the same prompt can cost more than the per-token ratio suggests. Measure on your own prompts.","source":"https://platform.claude.com/docs/en/about-claude/pricing"}],"notes":["Dates apply to the Claude API, Claude Platform on AWS and Microsoft Foundry. Amazon Bedrock and Google Cloud set their own retirement schedules."],"seen":"2026-10-08","provider":"anthropic","id":"claude-sonnet-4-5-20250929","shutdown":"2026-11-30","source":"https://platform.claude.com/docs/en/about-claude/model-deprecations","pricingSource":"https://platform.claude.com/docs/en/about-claude/pricing","pricingSeen":"2026-10-08","aliasSource":"https://platform.claude.com/docs/en/about-claude/models/model-ids-and-versions"},{"aliases":["gpt-5"],"kind":"model","announced":"2026-06-11","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[{"replacement":"gpt-5.6-sol","retiring":{"input":1.25,"output":10},"replacementPrice":{"input":4,"output":20},"ratio":{"input":3.2,"output":2},"note":"OpenAI lists gpt-5.6-sol's $4 / $20 as promotional pricing, available at least through 21 November 2026 (its model page, read 8 October 2026)."}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-5-2025-08-07","shutdown":"2026-12-11","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08","aliasSource":"https://developers.openai.com/api/docs/models/gpt-5"},{"aliases":["gpt-5-mini"],"kind":"model","announced":"2026-06-11","shutdownIsEarliest":false,"replacements":["gpt-5.6-terra"],"prices":[{"replacement":"gpt-5.6-terra","retiring":{"input":0.25,"output":2},"replacementPrice":{"input":2,"output":12},"ratio":{"input":8,"output":6}}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-5-mini-2025-08-07","shutdown":"2026-12-11","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08","aliasSource":"https://developers.openai.com/api/docs/models/gpt-5-mini"},{"aliases":["gpt-5-nano"],"kind":"model","announced":"2026-06-11","shutdownIsEarliest":false,"replacements":["gpt-5.6-luna"],"prices":[{"replacement":"gpt-5.6-luna","retiring":{"input":0.05,"output":0.4},"replacementPrice":{"input":0.2,"output":1.2},"ratio":{"input":4,"output":3}}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-5-nano-2025-08-07","shutdown":"2026-12-11","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08","aliasSource":"https://developers.openai.com/api/docs/models/gpt-5-nano"},{"aliases":["gpt-5-pro"],"kind":"model","announced":"2026-06-11","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[],"breaks":[{"short":"the replacement needs reasoning.mode set to pro","text":"OpenAI's replacement is gpt-5.6-sol with reasoning.mode set to pro, not the default mode.","source":"https://developers.openai.com/api/docs/deprecations"}],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-5-pro-2025-10-06","shutdown":"2026-12-11","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"No price for gpt-5.6-sol with reasoning.mode pro was listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given.","aliasSource":"https://developers.openai.com/api/docs/models/gpt-5-pro","replacementNote":"reasoning.mode: pro"},{"aliases":["o3"],"kind":"model","announced":"2026-06-11","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[{"replacement":"gpt-5.6-sol","retiring":{"input":2,"output":8},"replacementPrice":{"input":4,"output":20},"ratio":{"input":2,"output":2.5},"note":"OpenAI lists gpt-5.6-sol's $4 / $20 as promotional pricing, available at least through 21 November 2026 (its model page, read 8 October 2026)."}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"o3-2025-04-16","shutdown":"2026-12-11","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08","aliasSource":"https://developers.openai.com/api/docs/models/o3"},{"aliases":["o3-pro"],"kind":"model","announced":"2026-06-11","shutdownIsEarliest":false,"replacements":["gpt-5.6-sol"],"prices":[],"breaks":[{"short":"the replacement needs reasoning.mode set to pro","text":"OpenAI's replacement is gpt-5.6-sol with reasoning.mode set to pro, not the default mode.","source":"https://developers.openai.com/api/docs/deprecations"}],"notes":[],"seen":"2026-10-08","provider":"openai","id":"o3-pro-2025-06-10","shutdown":"2026-12-11","source":"https://developers.openai.com/api/docs/deprecations","priceNote":"No price for gpt-5.6-sol with reasoning.mode pro was listed on OpenAI's pricing page when checked on 8 October 2026, so no ratio is given.","aliasSource":"https://developers.openai.com/api/docs/models/o3-pro","replacementNote":"reasoning.mode: pro"},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":false,"replacements":["gemini-3.8-flash"],"prices":[{"replacement":"gemini-3.8-flash","retiring":{"input":1.5,"output":7.5},"replacementPrice":{"input":1.5,"output":7.5},"ratio":{"input":1,"output":1},"note":"Agent Platform lists both models at $0.75 / $3.75 through 31 December 2026 and $1.50 / $7.50 from 1 January 2027."}],"breaks":[],"notes":["A short-term availability model: Google posts a fixed retirement date at least 45 days ahead.","On the Gemini API (ai.google.dev) the same model ID had no shutdown date announced when checked on 8 October 2026."],"seen":"2026-10-08","provider":"vertex-ai","id":"gemini-3.7-flash","shutdown":"2027-01-28","source":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/model-versions","pricingSource":"https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing","pricingSeen":"2026-10-08"},{"aliases":[],"kind":"model","announced":"2026-10-01","shutdownIsEarliest":false,"replacements":["gpt-6-sol"],"prices":[{"replacement":"gpt-6-sol","retiring":{"input":1.25,"output":10},"replacementPrice":{"input":2,"output":10},"ratio":{"input":1.6,"output":1}}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-5.1","shutdown":"2027-04-01","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08"},{"aliases":[],"kind":"model","announced":"2026-10-01","shutdownIsEarliest":false,"replacements":["gpt-6-sol"],"prices":[{"replacement":"gpt-6-sol","retiring":{"input":1.75,"output":14},"replacementPrice":{"input":2,"output":10},"ratio":{"input":1.14,"output":0.714}}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-5.3-codex","shutdown":"2027-04-01","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08"},{"aliases":[],"kind":"model","announced":"2026-10-01","shutdownIsEarliest":false,"replacements":["gpt-6-luna"],"prices":[{"replacement":"gpt-6-luna","retiring":{"input":0.2,"output":1.25},"replacementPrice":{"input":0.1,"output":0.5},"ratio":{"input":0.5,"output":0.4}}],"breaks":[],"notes":[],"seen":"2026-10-08","provider":"openai","id":"gpt-5.4-nano","shutdown":"2027-04-01","source":"https://developers.openai.com/api/docs/deprecations","pricingSource":"https://developers.openai.com/api/docs/pricing","pricingSeen":"2026-10-08"},{"aliases":[],"kind":"model","announced":null,"shutdownIsEarliest":true,"replacements":["gemini-3.5-flash-lite"],"prices":[{"replacement":"gemini-3.5-flash-lite","retiring":{"input":0.25,"output":1.5},"replacementPrice":{"input":0.3,"output":2.5},"ratio":{"input":1.2,"output":1.67}}],"breaks":[],"notes":["Google lists this as the earliest possible shutdown date and says it will give the exact date with advance notice."],"seen":"2026-10-08","provider":"gemini-api","id":"gemini-3.1-flash-lite","shutdown":"2027-05-07","source":"https://ai.google.dev/gemini-api/docs/deprecations","pricingSource":"https://ai.google.dev/gemini-api/docs/pricing","pricingSeen":"2026-10-08"}]}