Claude Opus 4.8 fast mode error: what changes on 2026-11-30
As of 2026-11-05, Anthropic's release notes state that fast mode for Claude Opus 4.8 was deprecated on 2026-10-30 and will be removed on 2026-11-30; after that, a request to claude-opus-4-8 with speed: "fast" returns an error.
Updated 2026-11-05
One key for Claude, GPT and leading Chinese models, billed per token, up and running in minutes.
- Billed per token — live rates on /models
- Pay by card (Visa / Mastercard), Apple Pay, Google Pay or crypto
- Self-service: top up and activate instantly
Four takeaways
Removal date announced by Anthropic
Anthropic's 2026-10-30 release-notes entry deprecates fast mode for Opus 4.8 and names this day as the removal date; the Fast mode documentation gives the same date.
Which requests are affected
Only requests that set speed: "fast" (for Managed Agents, the speed field on the agent's model object); the docs describe speed: "fast" as the opt-in.
Where the docs send you
The release notes say to migrate to Claude Opus 5.5 to keep fast mode. Fast mode for Opus 5.5 left research preview on 2026-10-15: no access request and no fast-mode-2026-02-01 beta header needed.
Official fast mode price
Anthropic's price list (captured 2026-11-05): fast mode on Opus 5.5 costs $8 input and $40 output per million tokens; on Opus 4.8 and Opus 5 it is $10 / $50. QCode bills per token, with per-model rates on /models.
What happens to speed: fast after 2026-11-30
As of 2026-11-05 the official answer is plain: from 2026-11-30, requests to claude-opus-4-8 with speed: "fast" return an error, and nothing falls back to standard speed. Anthropic described the same handling (an error, no fallback) when it removed fast mode for Opus 4.7 on 2026-07-24, whereas the older Opus 4.6 ran at standard speed without an error. The model itself is not affected: claude-opus-4-8 is listed as Active, with a tentative retirement date no sooner than 2027-05-28. That leaves two fixes: drop the speed parameter and keep using Opus 4.8 at standard speed, or switch to claude-opus-5-5 and keep fast mode.
What the official pages say
The official wording sits in two places: the 2026-10-30 entry in the Anthropic platform release notes, and the availability section of the Fast mode documentation, which also links to the part of the Opus 5.5 migration guide on moving from Claude Opus 4.8. The 2026-10-15 entry adds that fast mode for Opus 5 and Opus 4.8 remains in research preview and tells you to keep sending the fast-mode-2026-02-01 beta header with fast mode requests to those models.
Timeline
Fast mode for Opus 5.5 leaves research preview: no access request and no fast-mode-2026-02-01 beta header needed (requests that still send it keep working). Fast mode for Opus 5 and Opus 4.8 stays in research preview.
Anthropic deprecates fast mode for Opus 4.8 in the release notes and announces 2026-11-30 as the removal date.
Removal date announced by Anthropic: fast mode for claude-opus-4-8 is removed on this day.
Confirmed vs common misconceptions
Confirmed (verifiable word for word)
The release notes and the Fast mode page agree: fast mode for Opus 4.8 is deprecated and will be removed, and keeping fast mode means migrating to Claude Opus 5.5. The precedent can be checked too: the 2026-07-24 release-notes entry says that after fast mode for Opus 4.7 was removed, claude-opus-4-7 requests with speed: "fast" return an error, while Opus 4.7 itself remains available at standard speed.
Common misconceptions
One: that Opus 4.8 itself goes offline on 2026-11-30. Only fast mode is removed, and the model table still lists claude-opus-4-8 as Active. Two: that requests quietly fall back to standard speed, or that the status code is already settled. Anthropic says the request returns an error and names no status code; for the exact error, go by the official release notes. Three: that changing the model string is enough. Opus 5.5 rejects thinking set to disabled and forced tool_choice, expects temperature, top_p and top_k to be omitted, and rejects a prefilled assistant turn.
Two ways to fix it
Drop speed and stay on Opus 4.8
Smallest change: delete speed: "fast" from the request and leave the rest; output speed returns to standard. Opus 4.8 still supports Priority Tier, which Opus 5.5 does not.
Switch to claude-opus-5-5 and keep fast mode
Change the model ID and leave speed: "fast" in place. Read the migration guide first: thinking is always on, forced tool_choice is rejected, and Priority Tier is not supported. The official standard price is $4 / $20 per million tokens, fast mode $8 / $40.
Batch-fixing your requests in three steps
① Search the whole repository for speed and claude-opus-4-8: in the Messages API check the speed field of each request, in Managed Agents the model object of each agent, and do not miss config files and hard-coded model names. ② For each call site pick one fix: delete speed: "fast", or set model to claude-opus-5-5 and check the migration-guide items: thinking (omit the field or use adaptive; disabled is rejected), effort, tool_choice (auto or none only), temperature, top_p and top_k (omit them), and prefill (not allowed). ③ Confirm the result in the response: the usage object has a speed field whose value is fast or standard.
Using it on QCode
As of 2026-11-05 the QCode /models page lists both claude-opus-4-8 and claude-opus-5-5; billing is per token with per-model rates on /models, and availability follows /models. Three steps: register and get your cr_ key, pay on the top-up page, then set ANTHROPIC_BASE_URL to https://api.qcode.cc/api. Not ready for a monthly plan? Pay-as-you-go Packs start at $4 for a one-time quota you use within 7 days; if you already have a Personal Plan, add a booster pack when you need more quota.
FAQ
What happens to my Opus 4.8 fast mode code after November 30?
From 2026-11-30, requests to claude-opus-4-8 with speed: "fast" return an error, with no fallback to standard speed. Until then fast mode is deprecated but not yet removed. Requests without speed are not affected by this announcement.
Does Opus 5.5 have fast mode?
Yes. The Fast mode page lists Opus 5.5 as available, with Opus 5 and Opus 4.8 in research preview. Anthropic says it gives up to 2.5x higher output tokens per second at premium pricing, $8 input and $40 output per million tokens, and you request it with speed: "fast" without a beta header.
What status code does the claude-opus-4-8 speed fast error return?
Anthropic says only that the request returns an error and gives no status code; for the exact error, go by the official release notes. The 429 or 529 mentioned in the Fast mode documentation is a different case: it applies when you exceed fast mode's rate limits or capacity.
Can I use claude-opus-4-8 and claude-opus-5-5 on QCode?
Yes: as of 2026-11-05 both models are listed on QCode's /models page, billed per token. The QCode endpoints page states that request bodies follow the corresponding official schema (Anthropic POST /v1/messages); adjust any requests that send speed: "fast" yourself before 2026-11-30, following the official notes.
Does Opus 4.8 itself get retired?
Not by this announcement. The model table lists claude-opus-4-8 as Active with a tentative retirement date no sooner than 2027-05-28; what is removed on 2026-11-30 is fast mode only.
How do I remove speed: fast from many call sites at once?
Search with grep -rn for speed and claude-opus-4-8, then either delete the speed argument or change model to claude-opus-5-5 and run through the migration guide's request checklist. Managed Agents set speed on the agent's model object, so check agent definitions as well as request code.
Sources
Anthropic platform.claude.com pages: release notes, Fast mode, Pricing, Model deprecations and the Claude Opus 5.5 migration guide. QCode pages: qcode.cc /models and /pay-as-you-go, and docs.qcode.cc endpoints and API paths. All captured 2026-11-05.
Send one request to check the fix
Payment methods and arrival times are on the top-up page; model availability follows /models.
Related reading
Claude Code one-time quota: no subscription, 7 days
A one-time quota with no subscription; whatever you buy is added to the same key. How that quota is shaped and its concurrency are on the QCode choosing-a-plan page.
Claude Opus 5.5 Guide (2026): $4/$20
Official Opus 5.5 pricing, the 1M context, cache read rates and how to call it on QCode; compare price and capability before you migrate.
Claude Adaptive Thinking & Effort 2026: Off Rules by Model
Which models can switch thinking off, the default effort levels and the budget_tokens change; a reference while moving to Opus 5.5.
Crypto payment guide
Nine coins, 56 networks, zero platform fee for AI coding API credits.
Official details on this page were checked on 2026-11-05 against the platform.claude.com pages captured that day, and QCode details against qcode.cc and docs.qcode.cc captured the same day. Official pages prevail, model availability follows /models, and official changes may happen without notice.