> ## Content Index
> Fetch the complete content index at: https://claudefox.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Reminder: Claude Sonnet 4.5 retires, and Opus 4.8 fast mode is removed, on November 30
- URL: https://claudefox.com/reminder-claude-sonnet-4-5-retires-and-opus-4-8-fast-mode-is-removed/
- Published: 2026-11-09T08:55:00.000Z
- Updated: 2026-11-20T02:32:02.000Z
- Description: Claude Sonnet 4.5 (claude-sonnet-4-5-20250929) retires on the Claude API on November 30, 2026, and fast mode for Claude Opus 4.8 is removed the same day. What breaks, how to find old calls, and what to change for Sonnet 5.5 and Opus 5.5.
- Author: hupicake
- Tags: News

*Updated November 20, 2026: Expanded with an audit checklist, Sonnet 5.5 migration changes and Claude Code steps.*

Two Claude deadlines fall on November 30, 2026\. Claude Sonnet 4.5 (claude-sonnet-4-5-20250929, plus its undated alias claude-sonnet-4-5) retires on the Claude API. On the same day, fast mode for Claude Opus 4.8 is removed. After that date, requests to the retired model fail, and Opus 4.8 requests that set speed to "fast" return an error.

## What changes on November 30

- Claude Sonnet 4.5: deprecated on September 30, 2026, and retires on November 30\. Anthropic recommends claude-sonnet-5-5 as the replacement.
- Fast mode on Claude Opus 4.8: deprecated on October 30 and removed on November 30\. Opus 4.8 itself stays active at standard speed; Anthropic lists its retirement as not sooner than May 28, 2027\. To keep fast mode, Anthropic points to Claude Opus 5.5.

These dates apply to Anthropic-operated platforms: the Claude API, Claude Platform on AWS, Claude Models on Google Cloud (Anthropic managed) and Microsoft Foundry. Amazon Bedrock and Google-managed Claude on Google Cloud set their own schedules, so check their model tables separately. Claude Code uses the Claude API, including when you sign in with a Claude plan, so the Anthropic dates apply there too, unless Claude Code is set up for another platform or a gateway, in which case that platform's dates apply.

## Find every place you still use them

Anthropic's deprecation page describes a quick audit: on the Usage page in Claude Console, click Export. The CSV breaks usage down by API key and by model, so any key still sending Sonnet 4.5 traffic shows up.

The export tells you which keys are affected, not where the calls live in your code. A repository search helps with that. This is our suggestion, not an Anthropic step: search for the model IDs, then for every way fast mode can be switched on (the speed field, the fastMode setting, /fast). The second search is noisy, so read the hits rather than trusting the count:

```
grep -rnE 'claude-sonnet-4-5|claude-opus-4-8' .
grep -rnE 'speed|fastMode' .
```

For Claude Code, Anthropic's checklist is to update Claude Code first, then check which model a session actually runs, even if you set none or use an alias such as sonnet, and change the old model everywhere it is set, subagent definitions included. Once Sonnet 4.5 retires, Claude Code on the Claude API shows an error beginning "There's an issue with the selected model", unless a fallback model chain moves it to the next model.

## Moving from Sonnet 4.5 to Sonnet 5.5

For Messages API code this is more than a string swap. If you use Claude Managed Agents, the guide says updating the model name is all you need. If you use Claude Code, the migration guide suggests letting it do the edit:

```
/claude-api migrate this project to claude-sonnet-5-5
```

The changes the guide lists for code coming from Sonnet 4.5:

- Thinking runs by default: a request with no thinking field uses adaptive thinking. Read content blocks by type, pass thinking blocks back unchanged in tool loops, and revisit max\_tokens, because thinking counts toward it and bills as output.
- To run without up-front thinking, send thinking with type "between\_tools"; "disabled" returns a 400 error. between\_tools takes no other fields, works only at low, medium and high effort (xhigh and max return a 400), and effort cannot change mid-conversation while it is on. Updates between tool calls still come back as thinking blocks.
- Assistant prefill is rejected. Use structured outputs for format control and move continuation text into the user message.
- Remove temperature, top\_p and top\_k, and replace thinking budgets with an effort level; Sonnet 5.5 defaults to high on the API.
- Forced tool choice (type any or tool) returns a 400 error. The guide says to use auto and mark the tool strict: true.
- Remove the interleaved-thinking-2025-05-14 and context-window beta headers, replace fine-grained-tool-streaming-2025-05-14 with eager\_input\_streaming: true per tool, and move output\_format to output\_config.format.
- Images use a higher-resolution tier, so a 2000×1500 image costs about 2.5 times as many tokens.

Sonnet 5.5 lists at $2 per million input tokens and $10 per million output, against $3 and $15 for Sonnet 4.5\. That is not a straight one-third saving: Claude 4.7 and later models use a tokenizer that turns the same text into roughly 30% more tokens, and Sonnet 5.5 thinks by default. Re-measure cost on real traffic.

## Keeping fast mode after Opus 4.8

On the API, switch the model to claude-opus-5-5 and keep speed set to "fast". Fast mode on Opus 5.5 is out of research preview, needs no beta header and costs $8 input and $40 output per million tokens, against $10 and $50 for Opus 4.8 fast mode. The migration guide lists three direct changes: thinking now runs on requests that omitted it, Priority Tier is not supported, and the minimum cacheable prompt drops from 1,024 to 512 tokens. It adds that its whole Opus 5 section applies, because Opus 4.8 still accepts disabled thinking, forced tool choice and the computer\_20251124 tool, returns text between tool calls as text blocks, and defaults to high effort. Check each against Opus 5.5.

In Claude Code there are two paths. To keep fast mode, update Claude Code and use /model to switch to Opus 5.5, which needs v2.1.280 or later; fast mode stays on after the switch. Fast mode in Claude Code is not available on Amazon Bedrock, Microsoft Foundry or Claude Platform on AWS, and subscription plans pay for it from usage credits. To stay on Opus 4.8 at standard speed, turn fast mode off:

```
/fast off
```

## Our take

As of November 20, ten days remain, which is enough if you start with the usage export rather than with memory, then search the code for model IDs and fast-mode settings. Make the model ID a config value in one place, run your evals on Sonnet 5.5 or Opus 5.5 before the 30th, and the next deprecation becomes a one-line change.

Source: [Model deprecations (Claude Platform Docs)](https://platform.claude.com/docs/en/about-claude/model-deprecations?ref=claudefox.com), [Claude Platform release notes](https://platform.claude.com/docs/en/release-notes/overview?ref=claudefox.com), [Migrating to Claude Sonnet 5.5 (Claude Platform Docs)](https://platform.claude.com/docs/en/models/sonnet-5-5/migration-guide?ref=claudefox.com), [Migrating to Claude Opus 5.5 (Claude Platform Docs)](https://platform.claude.com/docs/en/models/opus-5-5/migration-guide?ref=claudefox.com), [Speed up responses with fast mode (Claude Code Docs)](https://code.claude.com/docs/en/fast-mode?ref=claudefox.com), [Pricing (Claude Platform Docs)](https://platform.claude.com/docs/en/about-claude/pricing?ref=claudefox.com)

This story first appeared in the [claudefox daily briefing for November 9, 2026](https://claudefox.com/daily-2026-11-09/).