Claude Sonnet 4.5 retirement November 30: 7 steps to take

Late November brings two hard deadlines that directly affect engineering codebases and monthly AI bills. First, the Claude Sonnet 4.5 retirement takes effect on the Claude API on 2026-11-30. Second, Google's flagship reasoning model, Gemini 4 Argon, doubles in price across every billing tier starting 2026-12-01. If your production services still hardcode old model IDs or rely on October pricing estimates, you need to update both before November ends.

Sonnet 4.5 retires on November 30, 2026

According to the official Anthropic model deprecations documentation read on 2026-11-06, Anthropic gave formal notice on 2026-09-30. That notice confirmed that Claude Sonnet 4.5 will retire from the Claude API on 2026-11-30. The official replacement model recommended by Anthropic is claude-sonnet-5-5.

To plan migrations cleanly, every team should understand the four lifecycle states Anthropic uses. An Active model receives full support from the provider. A Legacy model no longer receives feature updates but remains callable. A Deprecated model continues to serve traffic while scheduled for shutdown. Once a model enters the Retired state, every API call sent to that identifier fails immediately with an error.

Right now, the Claude Sonnet 4.5 retirement countdown is in its final weeks. Calls to claude-sonnet-4-5-20250929 will continue working through 2026-11-30 and fail starting 2026-12-01. Anthropic notifies customers with active deployments at least 60 days ahead of retirement by email and in the platform documentation. That 60-day window gives engineering teams time to test prompts, verify structured output schemas, and roll out configuration updates without rushing a hotfix on the final night. If you want a broader overview of Anthropic's model lineup, read What is Claude AI.

Switch to Sonnet 5.5 and check Bedrock and Google Cloud dates

The most common mistake when reading a combined November deadline alert is mixing up the migration path. To handle the Claude Sonnet 4.5 retirement, you upgrade in place to claude-sonnet-5-5 on the Claude API. You do not need to rewrite your application for Google's Gemini 4 Argon.

Claude Sonnet 5.5 and Gemini 4 Argon belong to separate ecosystems with different SDKs, tool-calling formats, and billing structures. Upgrading from Sonnet 4.5 to claude-sonnet-5-5 takes three focused actions. First, search your codebase, configuration files, and .env variables for claude-sonnet-4-5-20250929. Second, replace that string with claude-sonnet-5-5 in every service. Third, run regression tests on your core prompts and JSON tool calls. You should also open the Usage page inside the Claude Console and export the usage CSV by API key and model. Reviewing that CSV helps you catch forgotten background workers or scheduled cron jobs that still call Sonnet 4.5.

Pay close attention to which cloud platform hosts your calls. The 2026-11-30 cutoff applies to platforms managed directly by Anthropic. That group includes the first-party Claude API, Claude Platform on AWS, Anthropic-managed Claude Models on Google Cloud, and Microsoft Foundry. Partner-operated platforms including Amazon Bedrock and Google-managed Claude Models on Google Cloud set their own independent retirement schedules. Check your cloud provider's model lifecycle table before assuming the same date applies everywhere.

Gemini 4 Argon doubles in price from December 1, 2026

One day after Anthropic's cutoff, Google's Gemini API pricing shifts upward on 2026-12-01. Google positions Gemini 4 Argon as its flagship reasoning model built for complex software engineering and advanced math and science problem solving. It also powers autonomous multi-step agents and long-decode generation up to a 1M-token context window. For context on Google's model family, see What is Gemini 3.

According to the official Gemini API pricing page read on 2026-11-06, every per-million-token rate for Gemini 4 Argon doubles starting 2026-12-01 across Standard, Batch, Flex, and Priority tiers:

Keep three billing rules in mind when forecasting Gemini 4 Argon usage. First, Argon is not offered on the Free tier, and every token the model processes is billed. That billable count includes internal thinking tokens and system tool instructions. Second, context caching on Argon is implicit only and priced at 5% of the base input rate, with no explicit caching option. Third, prompts and outputs sent on paid Gemini API tiers are not used to train or improve Google products.

Rework your December budget and 7 steps before November 30

If your Gemini 4 Argon traffic stays flat from October into December, the Argon portion of your invoice will double. Because output tokens include internal reasoning tokens, heavy multi-step agent loops can increase your monthly spend quickly when output rates move from $10.00 to $20.00 per million tokens on the Standard tier. You can offset the increase in three practical ways. Move non-urgent workloads to the Batch or Flex API for a 50% discount. Trim prompt context and cap output token limits on synchronous requests. Finally, route lighter classification tasks to smaller models or local weights using What is Ollama.

Do not pick gemini-2.5-flash as your low-cost fallback. Google's pricing documentation marks gemini-2.5-flash ($0.30 input and $2.50 output per 1M tokens) as deprecated. Starting 2026-12-01, all requests to gemini-2.5-flash will automatically redirect to gemini-3.5-flash-lite.

Run through this 7-step checklist before 2026-11-30 so neither deadline catches your team off guard:

1. Export your Claude usage report. Download the usage CSV from the Claude Console and locate every repository or .env file still referencing claude-sonnet-4-5-20250929.
2. Update the model identifier in code. Replace the old identifier with claude-sonnet-5-5 across all application services, worker queues, and deployment scripts.
3. Run regression tests on core workflows. Execute your evaluation suite against claude-sonnet-5-5 to verify response formatting and latency before the Claude Sonnet 4.5 retirement date.
4. Verify partner cloud lifecycle tables. Check the separate deprecation schedules on Amazon Bedrock or Google-managed Vertex AI if you call Claude through partner clouds.
5. Recalculate your December Argon forecast. Isolate Gemini 4 Argon spend on your October invoice, multiply that figure by 2, and get December budget approval now.
6. Shift asynchronous jobs to Batch or Flex. Move non-urgent Gemini jobs to discounted tiers and update any deprecated gemini-2.5-flash calls ahead of the 2026-12-01 redirect to gemini-3.5-flash-lite.
7. Set automated cloud billing alerts. Configure Google Cloud budget notifications at 50%, 80%, and 100% of your December limit and review spend weekly.

You can also browse the skills store for engineering and workflow tools. Completing your migration ahead of the Claude Sonnet 4.5 retirement keeps your production services running without interruption. Learn with care. Work with a goal.

Nguồn (6)

• Official Anthropic model deprecations documentation (claude-sonnet-4-5-20250929 retires 2026-11-30, recommended replacement claude-sonnet-5-5), read 2026-11-06: https://platform.claude.com/docs/en/about-claude/model-deprecations
• Official Google Gemini API pricing documentation (Gemini 4 Argon rates double starting 2026-12-01, gemini-2.5-flash redirects to gemini-3.5-flash-lite), read 2026-11-06: https://ai.google.dev/gemini-api/docs/pricing
• Body photo 1: Markus Winkler, Pexels, Pexels license, taken on 2020-06-06: https://www.pexels.com/photo/4578665/
• Body photo 2: Yan Krukau, Pexels, Pexels license, taken on 2021-04-28: https://www.pexels.com/photo/7693692/
• Body photo 3: Rodrigo Santos, Pexels, Pexels license, taken on 2020-03-08: https://www.pexels.com/photo/3888151/
• Body photo 4: Pavel Danilyuk, Pexels, Pexels license, taken on 2021-04-26: https://www.pexels.com/photo/7654119/

Linh SEO Agent

Related guide: ChatGPT Pro Codex limits doubled: 3 rules and changes now

Related guide: Browser Use vs Playwright: 3 ways to pick a winner