Claude Code Fast Mode on Opus 4.8: Removal Date and Fix

November 3, 2026 claude-code costs

Fast mode for Claude Opus 4.8 in Claude Code was deprecated on October 30, 2026 and is removed on November 30, 2026. After that date, a session that is still on Opus 4.8 with fast mode on gets its requests rejected by the API. You have two fixes, and both take one command: switch to Opus 5.5 (fast mode stays on), or run /fast off and stay on Opus 4.8 at standard speed.

Verified November 3, 2026 against the Claude Code fast mode page and the changelog entry for 2.1.316. We did not run a session that hits the removal, because it has not happened yet. Everything below is documentation, not a measurement.

Are you affected?

You are affected only if both of these are true at the same time:

The session's model is Opus 4.8.
Fast mode is on.

Run /fast in the session: it reports whether fast mode is on, and the ↯ icon beside the prompt shows while it is active. Run /status or check the model in /model for the model. If you never pinned Opus 4.8, you are most likely fine: from Claude Code 2.1.280 the opus alias and the fast mode default both point at Opus 5.5 (see Opus 5.5 in Claude Code).

The docs call out one trap: a new session can still start on Opus 4.8 even after you picked something else, for example when a pinned model comes from a settings file or an environment variable. Check those before assuming a /model switch stuck.

Fix 1: keep fast mode, move to Opus 5.5

/model

Pick Opus 5.5 (needs Claude Code 2.1.280 or later). Per the docs, fast mode stays on after the switch, and Claude Code shows a Fast mode ON confirmation. If a pin keeps pulling new sessions back to Opus 4.8, change the pin: a model key in settings or an ANTHROPIC_MODEL value.

Fix 2: stay on Opus 4.8 at standard speed

/fast off

You remain on Opus 4.8 and requests run at standard speed and standard pricing. Saved preference note: the saved fastMode setting is what new sessions start from, so set "fastMode": false in your user settings if you want that to hold across sessions.

What changes in cost

Fast mode is the same model with a different API configuration, priced higher per token. The docs list, per million tokens:

| Model | Fast input | Fast output | | - | - | - | | Opus 5.5 | $8 | $40 | | Opus 5 | $10 | $50 | | Opus 4.8 | $10 | $50 |

So moving fast mode users from Opus 4.8 to Opus 5.5 lowers the fast rate by 20 percent on both input and output (8 versus 10, 40 versus 50). That is arithmetic on the published rates, not a measured bill. Fast mode pricing is flat across the full 1M-token context.

Two cost details matter when you switch:

Fast mode and standard speed keep separate prompt caches. Turning fast mode on in the middle of a long conversation pays the fast uncached input price for the whole context once. A model switch rewrites the cache the same way, so do the switch at the start of a session where you can. Our prompt caching page covers what invalidates the cache.
On Pro, Max, Ultra, Team and Enterprise plans, fast mode bills from usage credits only and is not included in the plan's rate limits, even if you have plan usage left.

Where fast mode is not available

The fast mode page lists these limits, which also decide whether you can migrate at all:

Not on Amazon Bedrock, Google Cloud's Agent Platform, Microsoft Foundry or Claude Platform on AWS. Only the Anthropic Console API and subscription plans with usage credits.
Team and Enterprise: off by default, an Owner has to enable it.
Console organizations: Opus 5 and Opus 4.8 need fast mode access provisioned, and the Opus 5.5 page says it also depends on your usage tier. Without access, fast mode requests hit the fast mode rate limit and fall back to standard speed.
Managed settings can block it: fastMode: false, fastModePerSessionOptIn: true (refuses /fast on outside an interactive terminal) or an availableModels allowlist that excludes the Opus model.

If you run unattended jobs with claude -p, note that /fast in non-interactive mode only works in a session launched with fast mode in its settings, for example claude -p --settings '{"fastMode": true}' (Claude Code 2.1.205 or later). A headless job that relied on Opus 4.8 plus that flag is the case most likely to be forgotten until the API starts rejecting it.

FAQ

What happens on November 30 if I do nothing?

If the session is on Opus 4.8 with fast mode on, the API rejects its requests, per the fast mode page. Sessions on standard speed, or already on Opus 5.5, are not affected by this change.

Is fast mode for Opus 5.5 also a preview?

The docs describe fast mode for Opus 5.5 as available, and fast mode for Opus 5 and Opus 4.8 as a research preview. Opus 4.7 lost fast mode earlier: deprecated June 25, 2026 and removed July 24, 2026.

Will `/fast` switch my model for me?

If the current model does not support fast mode, turning fast mode on also switches you to Opus, and the new model is used from the next request. Switching to a model without support turns fast mode off. Switching back to a supported Opus turns it on again if your saved preference is on.

Can I turn fast mode off for a whole team?

Yes: set CLAUDE_CODE_DISABLE_FAST_MODE=1, or have an administrator set fastMode: false in managed settings. Team and Enterprise organizations also start with it disabled until an Owner enables it.

What did 2.1.316 say exactly?

The changelog entry says fast mode for Opus 4.8 is deprecated as of October 30, 2026 and removed November 30, 2026, and tells you to migrate to Opus 5.5.