Opus 4.8 Fast Mode Ends Nov 30: Migrate to Opus 5.5

TL;DR
Claude Code deprecated Opus 4.8 fast mode on Oct 30 and removes it Nov 30, 2026. Requests then error. Opus 5.5 fast costs $8/$40 per MTok - here is the fix.
Claude Code deprecated fast mode for Opus 4.8 on October 30, 2026, and removes it on November 30, 2026. After that date a fast mode request on Opus 4.8 returns an error instead of falling back to standard speed. The fix is to move to Opus 5.5, where fast mode is cheaper and needs no beta header.
Official Sources#
| Resource | What it confirms |
|---|---|
| Claude Code changelog, v2.1.316 | Deprecation date, removal date, error-not-fallback behavior |
| Claude Code fast mode docs | Supported models, per-MTok pricing, Opus 5.5 as default since v2.1.280 |
| Claude API fast mode docs | speed: "fast" usage, beta header, availability by model |
| Introducing Claude Opus 5.5 | Standard and fast mode pricing for Opus 5.5 |
| Opus 5.5 migration guide | What changes when you replace claude-opus-4-8 |
What the changelog says#
The v2.1.316 release notes carry one line that matters here: fast mode for Opus 4.8 is deprecated as of October 30, 2026 and will be removed on November 30, 2026. Fast mode requests on Opus 4.8 then "return an error and do not fall back to standard Opus 4.8", and the notes tell you to "migrate to Opus 5.5 to keep the speedup". The Claude API fast mode docs carry the same removal date for claude-opus-4-8 requests with speed: "fast".
The same release adds CLAUDE_CODE_SUBAGENT_EFFORT, which sets the effort subagents run at instead of inheriting the session's. That is not part of this migration, but it matters if you tune cost per task.
The angle: it is a silent-failure deadline, not a price change#
Most deprecations hurt on the day you read about them. This one hurts on a date, and only for people who pinned the combination. Interactive users will notice a toggle behaving differently. A script or a CI job that sets Opus 4.8 and speed: "fast" will simply start failing on December 1 with nothing to fall back on.
Who is exposed:
- Interactive Claude Code users who pressed
/fastwhile on Opus 4.8. Claude Code has defaulted fast mode to Opus 5.5 since v2.1.280, so most up-to-date installs are already fine. Check yours. - API callers sending
speed: "fast"withmodel="claude-opus-4-8"and thefast-mode-2026-02-01beta header. - Agent harnesses and configs that hard-code the model ID next to the fast flag.
What to do before November 30#
- Update Claude Code, then run
/fastand look at which model it reports. Per the docs, if your current model does not support fast mode, Claude Code switches to Opus. - Search your repos and CI for the pair. A quick grep:
grep -rn "claude-opus-4-8" . | grep -i fast, plus any wrapper that setsspeed. - Swap the model ID to
claude-opus-5-5. In the API the call keeps the same shape:
response = client.messages.create(
model="claude-opus-5-5",
max_tokens=4096,
speed="fast",
messages=[{"role": "user", "content": "..."}],
)
print(response.usage.speed) # "fast" or "standard"
- Check
usage.speedin the response. The API docs say aspeed: "fast"request on an unsupported model errors, and a successful fast request reports"fast". - Work through the Opus 5.5 migration guide for the non-fast changes below before you flip production traffic.
What changes besides the speed flag#
From the migration guide, going from claude-opus-4-8 to claude-opus-5-5:
- A request with no
thinkingfield now runs with thinking, and thinking cannot be disabled. Thinking tokens bill as output, so re-checkmax_tokensand lowereffortwhere you want less. temperature,top_pandtop_kare removed - omit them.- Priority Tier is not supported on Opus 5.5. The fast mode docs also say fast mode is not available with a Priority Tier commitment, so plan capacity separately if you have one.
- The minimum cacheable prompt drops from 1,024 to 512 tokens, so short prompts can start caching.
- Fast mode on Opus 5.5 needs no beta header. Opus 4.8 and Opus 5 still do.
Cost: the migration is a small discount#
| Model | Fast mode input | Fast mode output |
|---|---|---|
| Opus 5.5 | $8 / MTok | $40 / MTok |
| Opus 5 and Opus 4.8 | $10 / MTok | $50 / MTok |
That is 20% less on both sides. Anthropic also lists Opus 5.5 standard speed at $4 input and $20 output per million tokens and says it generates output more than 30% faster than Opus 5 - both vendor-reported. Two traps from the Claude Code docs carry over: the first time you enable fast mode in a conversation you pay the full uncached input price for the whole context, and fast and standard speed keep separate caches. On subscription plans fast mode bills as usage credits, not the plan's included limits. For the full cost model, see Claude Code fast mode: is 2.5x speed worth 2x price.
What people are actually saying#
The conversation so far is thin and mostly informational. SmartScope's write-up of the changelog entry (published November 1) correctly lists the dates and the error-not-fallback behavior, and tells readers to switch to Opus 5.5 before the 30th. It also flagged Opus 5.5 fast pricing as not yet confirmed at the time - the current Claude Code docs and Anthropic's launch page now both list $8/$40. A video titled "Fast mode on Opus 4.8 ends after November 30" in a Claude Code news roundup shows the topic is circulating, but I did not find a large Hacker News or Reddit thread to quote; the Hacker News feed was not reachable from this run.
The counter-case is simple: if you never used /fast, or you already run Opus 5.5, there is nothing to do. The notice says nothing about retiring standard-speed Opus 4.8, so do not conflate the two. Another fair objection is that Opus 5.5 is a different model, not a drop-in; the migration guide above lists real behavior changes, so test before you move production traffic. For where Opus 5.5 fits against its siblings, see Opus 5 vs Opus 4.8 vs Fable 5, and for day-to-day use see the Opus 5.5 Claude Code playbook.
FAQ#
When does Opus 4.8 fast mode stop working?#
Fast mode for Opus 4.8 was deprecated on October 30, 2026 and is removed on November 30, 2026. After removal, fast mode requests on Opus 4.8 return an error and do not fall back to standard Opus 4.8, per the Claude Code v2.1.316 notes.
Do I need to do anything if I never use /fast?#
No. The deprecation only affects fast mode requests on Opus 4.8. Standard-speed Opus 4.8 is not mentioned in the deprecation notice.
What does fast mode cost on Opus 5.5?#
Anthropic lists Opus 5.5 fast mode at $8 per million input tokens and $40 per million output tokens, against $10 and $50 on Opus 5 and Opus 4.8.
Sources#
| Source | URL |
|---|---|
| Claude Code releases (v2.1.316) | https://github.com/anthropics/claude-code/releases |
| Claude Code fast mode docs | https://code.claude.com/docs/en/fast-mode |
| Claude API fast mode docs | https://platform.claude.com/docs/en/build-with-claude/fast-mode |
| Introducing Claude Opus 5.5 | https://www.anthropic.com/claude-opus-5-5 |
| Opus 5.5 migration guide | https://platform.claude.com/docs/en/models/opus-5-5/migration-guide |
| SmartScope on the removal | https://smartscope.blog/en/blog/claude-code-opus-4-8-fast-mode-removed-2026/ |
Last updated: November 2, 2026
Continue Reading#
- Claude Code Fast Mode: Is 2.5x Speed Worth 2x Price? - the full cost model, first-enable charge and when speed pays off
- Claude Opus 5.5 Developer Guide: Pricing, API, Claude Code - everything else that shipped with Opus 5.5
- Opus 5.5 in Claude Code: The Playbook and Where It Breaks - effort levels and failure modes on the new default
- Claude Opus 5 vs Opus 4.8 vs Fable 5 - which Opus to pin if you are not moving yet
Get the next deep dive like this in your inbox
One email a week on claude-code and the rest of the AI dev stack. Free.
Read next on Claude Code
Claude Code Fast Mode: Is 2.5x Speed Worth 2x Price?
Claude Code fast mode pricing: $8/$40 per MTok on Opus 5.5, the first-enable context charge, separate rate limit pools, and when 2.5x speed pays off.
8 min readClaude Opus 5.5 Developer Guide: Pricing, API, Claude Code
Claude Opus 5.5 costs $4/$20 per million tokens with $0.20 cache reads. SDK examples, Claude Code setup, and when to pick it over Sonnet 5.5 or Haiku 5.5.
11 min readOpus 5.5 in Claude Code: The Playbook and Where It Breaks
Opus 5.5 in Claude Code: what each playbook tip changes, from naming the finish line to CLAUDE.md stop rules, and where Hacker News says long runs go wrong.
6 min readTechnical content at the intersection of AI and development. Building with AI agents, Claude Code, and modern dev tools - then showing you exactly how it works.







