AI Model Deprecations & Shutdown Schedule
Track official model retirements, shutdown countdown deadlines, and recommended migration alternatives before production pipelines break.
OpenAI Fine-tuning Closed to New Users; Pricing Page Restructured
OpenAI's pricing page now states fine-tuning is closed to new users, with existing users able to train for the coming months and inference continuing until base models are deprecated. The page also drops Azure/Bedrock pricing links and reorganizes use-case navigation.
OpenAI winds down fine-tuning platform; pricing page nav updated
Pricing page diff shows removal of fine-tuning references and a notice that OpenAI is winding down the fine-tuning platform, closed to new users. Existing users retain access. Minor nav/JS changes are cosmetic.
OpenAI winds down fine-tuning platform for new users
OpenAI is discontinuing its fine-tuning platform; new users can no longer access it, while existing users retain access temporarily. This requires affected customers to plan migration away from fine-tuned models.
OpenAI pricing page rebuild; gpt-3.5-turbo row now marked legacy
The pricing table was re-rendered with a new build hash and the gpt-3.5-turbo row now carries a 'Legacy' label. No numeric price changes are visible in the diff, but the legacy tag signals eventual sunset of the model.
Veo 3.1 Preview models deprecated; migrate to Gemini Omni Flash
Google removed Veo 3.1 preview models (including veo-3.1-lite-generate-preview) from the pricing page and added a deprecation notice: shutdown on October 22, 2026. Users must migrate to Gemini Omni Flash or GA models to avoid disruption.
Veo 3.1 Preview models deprecated, shutdown Oct 22, 2026
Google deprecates veo-3.1-generate-preview, veo-3.1-fast-generate-preview, and veo-3.1-lite-generate-preview with shutdown on Oct 22, 2026. Users must migrate to Gemini Omni Flash or GA models via Gemini Enterprise Agent Platform.
DeepSeek retires legacy v4-flash models, routes to V4.1-Flash-1017
Legacy model names deepseek-v4-flash and deepseek-v4-flash-vision-exp are now retired; requests are silently served by DeepSeek-V4.1-Flash-1017 and billed at Flash pricing. Users relying on the old model behavior should migrate to deepseek-flash.
OpenAI winds down fine-tuning platform for new users
OpenAI is sunsetting its fine-tuning platform: new users can no longer access it, though existing users can still create training jobs for a few more months. Fine-tuned models remain available for inference until their base models are deprecated.
OpenAI winds down fine-tuning platform for new users
OpenAI is closing the fine-tuning platform to new users; existing users can still create training jobs for the coming months, and fine-tuned models remain available for inference until their base models are deprecated.
Gemini 3.7 Flash Removed from Pricing Page
The Gemini 3.7 Flash model entry has been removed from Google's AI pricing page, suggesting the model may be deprecated, renamed, or replaced. Users relying on gemini-3.7-flash should verify availability and plan for potential migration.
Claude Haiku 4.5 removed from Anthropic models documentation
The Claude Haiku 4.5 entry has been removed from Anthropic's models documentation page. This may indicate a deprecation, retirement, or documentation restructuring, but the diff provides no replacement or sunset details.
Gemini 2.5 Flash Image deprecation notice on pricing page
Google's pricing page now warns that Gemini 2.5 Flash Image (gemini-2.5-flash-image) is deprecated and will shut down on October 2, 2026, directing users to migrate to Gemini 3.1 Flash Image or Gemini 3.1 Flash Lite Image to avoid service disruption.
DeepSeek V4 Pro retirement reversed; API service continues
DeepSeek has reversed its plan to retire V4 Pro on September 14, 2026. The model will remain available via API with unchanged billing, instead of being routed to V4.1 Flash. Users relying on V4 Pro can continue without migration.
Gemini 3.7 Flash replaced by Gemini 3.8 Flash on pricing page
Google's pricing page now lists Gemini 3.8 Flash as the current model, removing Gemini 3.7 Flash references. The new model is positioned for long-horizon software engineering and autonomous agents. No pricing figures changed in this diff, but the model lineup shift signals potential deprecation of 3
Anthropic introduces Claude Fable 5.1, replacing Fable 5
Anthropic's model overview now lists Claude Fable 5.1 as the top-tier model for demanding reasoning and long-horizon agentic work, replacing Claude Fable 5. The new model has a knowledge cutoff of Jun 2026 (vs Jan 2026 for Fable 5). API ID is claude-fable-5-1. This is a new model launch, not a break
Claude Opus 4.8 replaced by Claude Opus 5 in model table
Anthropic updated the models overview page, replacing Claude Opus 4.8 with Claude Opus 5 as the recommended model for complex agentic coding and enterprise work. The new model has a May 2026 knowledge cutoff (vs Jan 2026 for Opus 4.8). Users should migrate to the new model ID claude-opus-5.
Looking for live rates and all active models?
Browse verified token rates across 28 providers and 1,400+ active models.