AI and security briefsModels
How do I migrate from Claude Opus 5.5 to Fable 5.2?
Published Updated Editorial policy
Change the model ID to claude-fable-5-2, but check four things first. Fable 5.2 needs 30-day data retention, enforces the prefix check on every account, defaults to high effort where Opus 5.5 uses medium, and has its own lower rate limit. Everything else in an Opus 5.5 request carries over.
Who should act: Teams that call Claude Opus 5.5 from their own code, a router or a fallback path and want to move some or all of that traffic to Claude Fable 5.2.
Update log
- : Published.
What changed
Anthropic released Claude Fable 5.2 on 9 November 2026 and published a migration path from Claude Opus 5.5 the same day (Anthropic migration guide, 9 November). Both use the same Messages API, both always think adaptively and both reject forced tool choice, so those parts carry over.
The guide lists nine differences. These are the ones that can break a request that works on Opus 5.5:
- Data retention. Fable 5.2 needs 30-day retention. Under zero data retention every request returns a 400 unless Anthropic has authorised it.
- The prefix check. Opus 5.5 enforces it for accounts created on or after 31 August 2026. Fable 5.2 enforces it for every account. Code that edits
system,toolsor earlier messages fails with the error "The block is bound to a different conversation". - Structured outputs under customer-managed keys. In a CMEK organisation,
strict: trueandoutput_config.formatare not available on Fable 5.2. - The advisor tool. A Fable 5.2 executor accepts only Fable 5.2 or Mythos 5.2 as its advisor. Any other returns a 400.
Three more change cost and behaviour without an error. The default effort is high on Fable 5.2 and medium on Opus 5.5. The rate limit is separate from Opus 5.5's, with lower token limits at each tier. And Opus 5.5 cannot read Fable 5.2's thinking blocks, so a fallback from Fable 5.2 to Opus 5.5 runs without the earlier reasoning (Anthropic what's new page, 9 November).
Who it affects
- Teams that build the
messagesarray themselves. Claude Code, Managed Agents and the Agent SDK keep history append-only, and Anthropic says nothing changes for them. - Organisations on zero data retention. Configure retention for a workspace or talk to the account team before anything else.
- Routers and fallback chains. Moving from Opus 5.5 onto Fable 5.2 keeps the reasoning. Moving back loses it, and the first turn after the switch costs more.
- Anyone who left effort unset. The same request now runs one level higher, at twice the token price.
What to do now
- Find what will break.
grep -rnE "claude-opus-5-5|tool_choice|computer_20251124|output_config|advisor|speed" src/ - Check retention. In the Console, open the workspace's data retention setting and confirm it is 30 days for the workspace that will call Fable 5.2.
- Send one request with the new ID and effort set explicitly.
curl https://api.anthropic.com/v1/messages -H "x-api-key: $ANTHROPIC_API_KEY" -H "anthropic-version: 2023-06-01" -H "content-type: application/json" -d '{"model":"claude-fable-5-2","max_tokens":8000,"output_config":{"effort":"medium"},"messages":[{"role":"user","content":"ping"}]}' - Test the prefix check. Replay a real multi-turn conversation from your logs. A 400 that names a different conversation means your code edits history. Fix the edit, and use the
drop_blocksetting only as a stopgap. - Compare both models' rate limits in the Console before sizing the first traffic share.
- If you use the advisor tool, set its model to
claude-fable-5-2in new conversations, because editingtoolsmid-conversation also fails the prefix check. - Re-run your evals at medium and high effort, then move 5 to 10% of traffic and keep Opus 5.5 as the fallback.
What to watch
- Cost per finished task. Fable 5.2 costs twice Opus 5.5 per token, so measure tasks completed, not tokens.
- Turns that stop early. Anthropic says Fable 5.2 sometimes ends a long task with a list of what remains, more often at low and medium effort.
- Refusals. Handle every
stop_detailscategory. Fable 5.2's permitted fallback targets are Opus 4.8 and Opus 5.
Questions people ask
Is Claude Fable 5.2 a drop-in replacement for Opus 5.5?
For most request shapes, yes. Data retention, the prefix check, structured outputs under customer-managed keys and the advisor pairing can still break a working integration.
Why do my requests fail on Fable 5.2 but work on Opus 5.5?
The two most likely causes are zero data retention, which Fable 5.2 does not allow without authorisation, and the prefix check, which Fable 5.2 enforces on every account. Older accounts never saw that check on Opus 5.5.
Does Fable 5.2 use the same effort setting as Opus 5.5?
It supports the same five levels, but the API default differs. Fable 5.2 defaults to high and Opus 5.5 to medium, so set effort explicitly before you compare cost or latency.
Can I fall back from Fable 5.2 to Opus 5.5?
Yes, but Opus 5.5 cannot read Fable 5.2's thinking blocks. The API drops them, the request succeeds and they are not billed. The turns after the switch run without the earlier reasoning.
Do I need to change anything in Claude Code?
No code change. Pick the model with /model claude-fable-5-2. Claude Code and the Claude apps default to medium effort.
Digiton's production review checks which model ID, effort, retention setting and fallback each service uses before a switch like this, and the checks are on the vibe coded app audit page.
How this brief was made: researched from the sources below, drafted with AI assistance and checked against each source on the dates shown. See the editorial policy.
Sources
- Claude Fable 5.2 and Mythos 5.2 migration guide, 9 November 2026, read 9 November 2026. https://platform.claude.com/docs/en/models/fable-5-2/migration-guide
- What's new in Claude Fable 5.2, 9 November 2026, read 9 November 2026. https://platform.claude.com/docs/en/models/fable-5-2/whats-new-fable-5-2
- Claude Fable 5.2 model page, 9 November 2026, read 9 November 2026. https://platform.claude.com/docs/en/models/fable-5-2/overview
- Claude Opus 5.5 model page, read 9 November 2026. https://platform.claude.com/docs/en/models/opus-5-5/overview
- Anthropic announcement of Claude Fable 5.2 and Claude Mythos 5.2, 9 November 2026, read 9 November 2026. https://www.anthropic.com/claude-fable-and-mythos-5-2
More briefs
How these briefs are sourced and corrected. All AI and security briefs
Building something that has to hold up in production?
Start with a free 25-minute scoping call. If it fits, the first paid step is a production review: five working days on your real repository, scored against ten layers, with a costed fix list.
Book a 25-minute scoping call