Claude removes Opus 4.8 fast mode, users should migrate to Opus 5.5
Claude announced that the “fast mode” for Opus 4.8 will be removed on November 30, 2026. After that date any request to claude‑opus‑4‑8 that includes speed: \"fast\" will return an error. The deprecation was first noted in the October 30, 2026 release notes and confirmed on October 31, 2026. Users of the Claude API, Claude Managed Agents, or the fast‑mode‑2026‑02‑01 beta header must switch to…
Key points
- Fast mode for Claude Opus 4.8 ends on November 30, 2026; requests with speed:\"fast\" will error
- Migration to Opus 5.5 requires changing model ID and removing or adjusting thinking, tool choice, temperature, top‑p, top‑k, and prefill settings
- Opus 5.5 fast mode costs $8 per million input tokens and $40 per million output tokens, lower than Opus 4.8 fast mode
The migration guide advises only changing the model ID and keeping the speed: \"fast\" flag, but several settings accepted by Opus 4.8 are rejected by Opus 5.5. These include disabling thinking, forced tool calls, explicit temperature/top‑p/top‑k, and pre‑fill messages. Effort defaults shift from high to medium, and thinking tokens are billed as output tokens. Priority Tier is not supported in Opus 5.5, and prompt‑caching minimum length drops from 1,024 to 512 tokens.
Pricing per 1 million tokens also changes: Opus 4.8 fast mode was $10 input / $50 output, while Opus 5.5 fast mode is $8 input / $40 output. Users should re‑measure cost and latency after migration and complete the switch before the November 30 deadline.
Model page: Claude Opus 5.5 →
The story so far
2 episodes →- Claude removes Opus 4.8 fast mode, users should migrate to Opus 5.5this story
Claude Opus 4.8 fast mode to be removed on November 30. What to fix when migrating to Opus 5.5
note.com · 31 October 2026
Loading the full article…
This text was published by note.com and written by HARu. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- AI News Summary 11/1 | Claude's Ultra 5-hour limit removed, 9 AI infrastructure companies breached, and more · 1 src
- Anthropic renames $200 Claude plan to Ultra and removes 5-hour limit · 1 src
- Krlntn builds neural network that boots PC from 512‑byte MBR · 1 src
- Author reviews Opus 5.5, dots and Muse AI secretaries · 1 src
- Amazon launches Alexa+ early access in Japan with generative AI for Prime members · 3 src
Comments
via GitHub Discussions