Coding playbooks · Reference table
Agent change deadlines: retirements and shutdowns
Written November 5, 2026
Every retirement, shutdown, and enforcement date that affects agent builders, in one countdown list — Sonnet 4.5, Agent Builder, Argon, and more.
Soonest first. Every date below comes from the vendor's own page, linked in the row. Where a vendor gives different dates per platform, the row says so — a date for one platform is not a date for all of them. Dates that have passed are removed as the list is updated.
The list
| Date | What changes | Who is affected | Source |
|---|---|---|---|
| November 9, 2026, 16:00–18:00 UTC | GitHub blocks pull_request_target workflows for two hours in public repos, unless an Actions event policy allows the trigger. First of two brownouts ahead of enforcement. | Public repos using pull_request_target without an event policy | GitHub changelog: pull_request_target blocked by default December 2 |
| November 16, 2026, 16:00–18:00 UTC | Second pull_request_target brownout, same two-hour window. | Public repos using pull_request_target without an event policy | GitHub changelog: pull_request_target blocked by default December 2 |
| November 30, 2026 | Claude Sonnet 4.5 (claude-sonnet-4-5-20250929) retires on Anthropic-operated platforms. Requests after retirement fail. Bedrock and Google-managed platforms set their own schedules — check theirs separately. | Anyone calling Sonnet 4.5 on the Claude API, Claude Platform on AWS, Google Cloud Anthropic-managed, or Microsoft Foundry | Anthropic model deprecations · Full playbook: Move off Claude Sonnet 4.5 before retirement |
| November 30, 2026 | Fast mode is removed for Claude Opus 4.8. After removal, requests to claude-opus-4-8 with speed: "fast" return an error and do not fall back. | Anyone using fast mode on Opus 4.8 — migrate to Opus 5.5 | Claude Platform release notes |
| November 30, 2026 | OpenAI shuts down Agent Builder, the Evals dashboard and API, and the reusable prompts (v1/prompts) API. Evals turned read-only on October 31. | Anyone using Agent Builder, Evals, or reusable prompts | OpenAI API deprecations · Full playbook: OpenAI November 30 shutdowns migration |
| November 30, 2026 | Free use of SWE-2 Flash ends. Cognition offered it free through November 30 in Devin Desktop and the CLI. | Anyone using SWE-2 Flash free in Devin | SWE-2 Flash announcement — Cognition |
| December 1, 2026 | Gemini 4 Argon doubles in price on the standard tier: $2.00 / $0.10 / $10.00 per million tokens becomes $4.00 / $0.20 / $20.00. Batch, Flex, and Priority tiers double too. | Anyone calling Gemini 4 Argon | Google Gemini API pricing · Full playbook: Gemini 4 Argon price doubling December 1 |
| December 1, 2026 | Gemini 2.5 Pro and 2.5 Flash are deprecated in AI Studio and no longer selectable. Requests auto-route instead of failing: 2.5 Pro goes to Gemini 3.8 Flash, billed at 3.8 Flash rates; 2.5 Flash goes to Gemini 3.5 Flash-Lite, billed at 3.5 Flash-Lite rates. Gemini 2.5 Flash-Lite is not deprecated. | Anyone calling Gemini 2.5 Pro or Flash in AI Studio | Google AI Studio changelog |
| December 1, 2026 | GPT Image models are removed: gpt-image-1-mini, gpt-image-1.5, and chatgpt-image-latest. | Anyone calling the GPT Image models | OpenAI API deprecations |
| December 2, 2026 | GitHub blocks pull_request_target by default in all public repos, unless an Actions event policy explicitly allows the trigger. Private and internal repos are unaffected. | Public repos using pull_request_target | GitHub changelog: pull_request_target blocked by default December 2 · Full playbook: pull_request_target December enforcement |
| December 11, 2026 | GPT-5 family and o3 snapshots are removed from the API: gpt-5, gpt-5-mini, gpt-5-nano, gpt-5-pro, o3, and o3-pro. | Anyone pinned to a GPT-5 or o3 snapshot | OpenAI API deprecations |
| January 14, 2027 | gpt-4o and gpt-4.1 snapshots are removed: gpt-4o-2024-08-06, gpt-4o-2024-11-20, gpt-4o-mini-2024-07-18, gpt-4.1-2025-04-14, and gpt-4.1-mini-2025-04-14. OpenAI names GPT-6 Luna as the replacement. | Anyone pinned to a gpt-4o or gpt-4.1 snapshot | OpenAI API deprecations |
How to use this list
Work down from the top. A retirement that fails requests needs a migration before the date; a reroute like the Gemini 2.5 change needs a billing check, because the replacement model bills at its own rates; a price doubling needs a budget check. For model prices across vendors, see the model price list.
Related tools
- Claude Code changes
- Model prices: what the major models cost
- Model specs: context, tools, and cutoffs in one table
- Open-weight model licences, in one list
Sources
- Anthropic model deprecations — Sonnet 4.5 retirement date and platform scope
- Claude Platform release notes — Opus 4.8 fast mode removal
- OpenAI API deprecations — Agent Builder, Evals, reusable prompts, GPT-5, o3, GPT Image, GPT-4o and GPT-4.1 removals
- SWE-2 Flash announcement — Cognition: free use through November 30
- Google Gemini API pricing — Gemini 4 Argon price doubling
- Google AI Studio changelog — Gemini 2.5 Pro and Flash deprecation and rerouting
- GitHub changelog: pull_request_target blocked by default December 2 — brownout windows and enforcement date
Back to: Coding playbooks hub · Also: Model price list · Move off Claude Sonnet 4.5 before retirement · OpenAI November 30 shutdowns migration