The short version
- Anthropic deprecated fast mode for Claude Opus 4.8 as of 30 October 2026 and will remove it on 30 November 2026.
- After removal, fast requests on Opus 4.8 return an error instead of quietly running at normal speed, so anything that depends on them can break.
- Anthropic says to migrate to Claude Opus 5.5, where fast mode is available and costs $8 per million input tokens and $40 per million output tokens.
- If nobody on your team turned fast mode on, nothing changes for you; if someone did, check their settings and any automated jobs this month.
On this page
If your team uses fast mode with Claude Opus 4.8, the clock is running. Anthropic deprecated that option as of 30 October 2026 and will remove it on 30 November 2026. After that date, a fast request on Opus 4.8 fails with an error. It does not fall back to normal speed. Anthropic's advice is to move to Opus 5.5.
This article explains what fast mode is, what changed, who is affected, and what to do in the next four weeks.
What is Claude fast mode?
Fast mode is a switch that makes Claude Opus answer faster, at a higher price per token. A token is a small chunk of text, roughly a short word or part of one, and AI usage is billed per million of them.
Anthropic's documentation says fast mode makes the model up to 2.5 times faster. On the API, the gain is in how quickly the answer is written out, not in how soon the first word appears. It is not a different model. It is the same Opus running with a configuration that favours speed over cost, and Anthropic says quality and abilities are identical.
In Claude Code, Anthropic's coding tool, the person at the keyboard turns fast mode on with a command called /fast, or by setting fastMode to true in their settings file. A small ↯ icon shows next to the prompt while it is on. Developers who call Claude directly through the API, which is how software talks to the model, ask for it by adding speed: "fast" to a request.
What changed on 30 October 2026?
The Claude Code changelog for version 2.1.316, released on 31 October 2026, says fast mode for Opus 4.8 was deprecated as of 30 October 2026 and will be removed on 30 November 2026. It adds that, after removal, fast requests on Opus 4.8 return an error and do not fall back to standard Opus 4.8. It tells readers to migrate to Opus 5.5 to keep the speedup.
Anthropic's API documentation says the same: requests to claude-opus-4-8 with fast speed will return an error after the removal date, and the way to keep fast mode is to move to Claude Opus 5.5.
Two details are worth noting:
- "Deprecated" means the feature still works today but is on its way out. The hard stop is 30 November.
- The failure is an error, not a slowdown. Work that used to run quickly will stop rather than carry on at normal speed, so an automated job that depends on fast mode can break.
One caution: Anthropic's Claude Code fast mode page, when we read it, still listed Opus 4.8 as supported (in research preview) and did not mention the removal. The changelog and the API page are where the dates appear, so treat those as the record.
Who is affected?
Most teams are not affected. Fast mode is something a person has to turn on, and on Team and Enterprise plans an organisation Owner must enable it first. You are affected if any of these is true:
- An engineer on your team uses Claude Code on Opus 4.8 and has turned fast mode on. By default, a fast mode choice made in an interactive session is remembered across sessions, so it may be on without anyone thinking about it.
- A script, tool or product your team built calls the Claude API with fast speed on Opus 4.8.
- A settings file shared by the team turns
fastModeon and uses Opus 4.8.
If you are a founder or manager who does not touch these settings, the practical question is simply: did anyone on the team switch this on? A one-line message to your engineers answers it.
What should we do before 30 November?
The fix is a move to a newer model, not a rebuild. Here is a short plan:
- Ask who uses fast mode. In Claude Code, the
↯icon next to the prompt shows it is on, and thefastModeline in the settings file shows the saved choice. - Move to Opus 5.5. Anthropic's documentation says Opus 5.5 is the fast mode default in Claude Code from version 2.1.280 onward. On the API, fast mode on Opus 5.5 is listed as available, while Opus 5 and Opus 4.8 are research previews.
- Check automated jobs. Anything that sends fast requests to Opus 4.8 needs the model name changed. Test it before the deadline, not after it fails.
- Re-check your budget. The price is different on the new model (see below).
- Write down the date. Put 30 November in the team calendar.
Anthropic also has a guide for migrating from Opus 4.8 to Opus 5.5. Your engineers should read it before changing model names in anything that runs unattended, since a new model can behave differently on the same request.
What does fast mode cost on Opus 5.5?
Per million tokens, Anthropic lists fast mode at these prices:
- Opus 5.5: $8 for input and $40 for output.
- Opus 5 and Opus 4.8: $10 for input and $50 for output.
So on paper, the replacement is cheaper per token than Opus 4.8 fast mode. Other costs and conditions still apply:
- On Claude subscription plans (Pro, Max, Ultra, Team, Enterprise), fast mode is paid for from usage credits only. It is not covered by the plan's included usage, even if you have usage left. Usage credits must be switched on first.
- On Team and Enterprise plans, an Owner must enable fast mode for the organisation before anyone can use it.
- Teams that pay through the API need a paid account. For Opus 5.5, fast mode also depends on the account's usage tier, the level that sets how much you can send.
- The first time you switch fast mode on in a conversation, you pay the full fast price for everything already in that conversation. Fast and normal speed keep separate caches, so each later switch reads part or all of the conversation again at full price. Anthropic says to switch it on at the start of a session.
Is fast mode worth keeping at all?
Anthropic's own advice is that fast mode suits interactive work where waiting hurts: quick iteration, live debugging, deadline pressure. It suggests normal speed for long autonomous tasks, batch jobs and automated pipelines, where speed matters less and cost matters more. The API's Batch interface does not support fast mode at all.
That makes this deadline a good moment to ask a plain question. Is anyone on the team really sitting and waiting on answers, and is the extra spend buying them time that matters? If not, the cheapest way to handle the removal is to let fast mode go.
How does Atlas handle this?
Atlas does not yet do anything specific about fast mode or model retirements. Atlas is a team of AI engineers built on Claude Code, running on your own Claude plan, and it sets the model and effort for each engineer according to the role. It has no notice for when a mode or model your team relies on is about to be retired. That is coming to Atlas, built in, so it works without any setup on your side. Until then, check the fast mode setting in your own Claude Code and in any scripts you run yourselves, using the steps above. You can read more about how the team works on the Atlas home page.
Conclusion
The change is small but dated: fast mode on Opus 4.8 stops working on 30 November 2026, and it stops with an error. If nobody on your team uses fast mode, do nothing. If someone does, spend ten minutes this week finding out where, move those uses to Opus 5.5, and decide whether the speed is worth the credits.
Questions people ask
When does fast mode for Claude Opus 4.8 stop working?
Anthropic deprecated it as of 30 October 2026 and will remove it on 30 November 2026. After that, fast requests on Opus 4.8 return an error and do not fall back to standard speed.
What should I use instead of Opus 4.8 fast mode?
Anthropic says to migrate to Claude Opus 5.5, where fast mode is available. In Claude Code, Opus 5.5 is the fast mode default from version 2.1.280 onward.
How much does fast mode cost on Opus 5.5?
Anthropic lists $8 per million input tokens and $40 per million output tokens on Opus 5.5, against $10 and $50 on Opus 4.8. On subscription plans it is paid from usage credits, not the plan's included usage.
Do I need to do anything if I never turned on fast mode?
No. Fast mode is a switch someone has to turn on. Ask your engineers whether anyone did, and whether any script sends fast requests to Opus 4.8.
Sources
Researched and drafted with Claude, then checked against the sources above. Atlas is an independent product, not affiliated with Anthropic or OpenAI. Claude and Claude Code are trademarks of Anthropic; Codex is a product of OpenAI.