Claude Code Fast Mode: Cost and Switching

Claude Code fast mode is not a separate model but a faster way of running Claude Opus: the same model with a different API configuration, up to 2.5x faster with identical quality, according to Anthropic. You turn it on with /fast. It costs twice the standard rate: $8 input and $40 output per million tokens on Opus 5.5, and $10 and $50 on Opus 5 and Opus 4.8. On Pro, Max, Ultra, Team and Enterprise, fast mode is paid for from usage credits only, never from your plan's included usage. And according to Anthropic's notice to Team and Enterprise admins dated October 30, 2026, fast mode for Opus 4.8 is retiring in Claude Code: after November 30, 2026, those requests will fail. Its successor is fast mode on Opus 5.5.

I work with Claude Code every day and run 14 AI employees on it. This article explains fast mode based on Anthropic's docs as of October 30, 2026; I am not describing a hands-on test of fast mode of my own here. Everything about Claude is collected on my Claude page.

How to turn on Claude Code fast mode

In the terminal you have two options:

1. With /fast: run the command, press Space to toggle, Enter to confirm. You'll see "Fast mode ON" and a small ↯ next to the prompt. Run /fast again at any time to check whether it is on.
2. In your settings: add "fastMode": true to your user settings file.

By default, the toggle sticks across sessions. If your model doesn't support fast mode, turning it on switches you to Opus (Opus 5.5 from v2.1.280). Switch to a model without fast mode, such as Opus 4.7, and it turns off.

Elsewhere:

• VS Code: the extension offers a Toggle fast mode command whenever the selected model supports it.
• Cloud sessions: /fast on applies to that session only; the browser also has a switch in the model menu. The environment needs Claude Code v2.1.271 or later. More on cloud sessions in Claude Code on the Web: Cloud Sessions.
• Non-interactive with -p: only when the session starts with fast mode in --settings, for example claude -p --settings '{"fastMode": true}', from v2.1.205.

What fast mode costs

Prices per million tokens (MTok) according to Anthropic, with the standard API rate for comparison:

Fast mode pricing is flat across the full 1M token context window. On Opus 5 and Opus 4.8 it is a research preview, so pricing and availability may change. Opus 4.7 lost fast mode on July 24, 2026. All current models are compared in Claude Models: Opus, Sonnet, Haiku, Fable.

Who pays depends on how you signed in, which /status shows you:

• Pro, Max and Ultra: fast mode draws from your usage credits, even if your plan still has usage left. You have to turn usage credits on under Settings > Usage on claude.ai first; until then, /fast reports "Fast mode requires usage credits".
• Team and Enterprise: your organization pays from its usage credits. /usage shows your own spend.
• Claude Console (API key): billed per token along with the rest of your API usage, and not available on the free Evaluation plan. An admin enables fast mode in the Console's Claude Code preferences. Opus 5 and Opus 4.8 also need fast mode access provisioned for the organization, and on Opus 5.5 it depends on the organization's usage tier. How API billing works in general is covered in Claude API Key: Setup, Pricing, First Call.

The context trap: turn it on early

The first time you enable fast mode in a conversation, the docs say you pay the full fast mode uncached input price for the entire conversation so far. Because fast mode and standard speed keep separate caches, every later switch reads part or all of it uncached again.

A quick calculation on Opus 5.5: switching on at 20,000 tokens of context costs $0.16 for the input of that first request alone. At 400,000 tokens it is $3.20; at standard speed, the same context read from cache would have cost $0.08. That is why Anthropic recommends enabling fast mode at the start of a session rather than halfway through.

Fast mode for Opus 4.8 ends November 30: how to switch

On October 30, 2026, the public docs still list fast mode for Opus 4.8 as a research preview. According to Anthropic's notice to Team and Enterprise admins from the same day, it is retiring in Claude Code: after November 30, requests with fast mode on will fail. Opus 4.8 itself stays; Anthropic's model deprecations page lists its retirement as not sooner than May 28, 2027.

What to do before November 30:

1. Check your version with claude --version. Fast mode on Opus 5.5 needs v2.1.280 or later.
2. To keep the speed, select Opus 5.5 with /model and run /fast. As a side effect, the price drops from $10/$50 to $8/$40 per million tokens.
3. To stay on Opus 4.8, run /fast off. Opus 4.8 then keeps running at standard speed.
4. Update pinned setups. Per the notice, workflows and configurations pinned to fast mode on Opus 4.8 should move to Opus 5.5. Models are set in places like the model setting, --model, ANTHROPIC_MODEL, ANTHROPIC_DEFAULT_OPUS_MODEL (what the opus alias points to), or subagent and skill frontmatter; fast mode in fastMode or in --settings for claude -p.
5. Check your allowlist. If your organization restricts models with availableModels, Opus 5.5 has to be on it. Otherwise /fast refuses with "is not in your organization's allowed models".

When fast mode is worth it, and when it isn't

Anthropic recommends fast mode for interactive work where waiting costs you more than tokens do:

• rapid iteration on code changes
• live debugging sessions
• time-sensitive work with tight deadlines

The docs recommend standard mode for long autonomous tasks, batch processing and CI/CD pipelines, and cost-sensitive work. If Claude Code runs overnight or on a schedule, nobody is waiting, and fast mode would only double the bill.

A lower effort level also speeds things up, differently: fast mode keeps quality and costs more, while lower effort cuts thinking time and can hurt quality on complex tasks. For simple tasks, combine both.

Fast mode has its own rate limits. Hit them and Claude Code falls back to standard speed and pricing, the ↯ turns gray, and fast mode returns after the cooldown.

Fast mode on Team and Enterprise

Fast mode is off by default for Team and Enterprise organizations. Two switches need to be on:

• Fast mode itself: an Owner enables it under Organization settings > Claude Code. Until then, /fast says "Fast mode has been disabled by your organization".
• Usage credits: a member with billing access turns them on under Organization settings > Usage; everyone else can ask the admins with /usage-credits.

To keep costs in check, fastModePerSessionOptIn: true makes every session start with fast mode off until someone runs /fast. The docs suggest it for organizations where people run several sessions at once. When Owners deploy it through managed settings, /fast on only works in an interactive terminal session, not in VS Code, cloud sessions or with -p. To switch fast mode off entirely, set fastMode: false in managed settings or CLAUDE_CODE_DISABLE_FAST_MODE=1.

According to the docs, fast mode in Claude Code is not available through Amazon Bedrock, Google Cloud's Agent Platform, Microsoft Foundry or Claude Platform on AWS.

Frequently asked questions

Is fast mode a different model?

No. It is the same Opus model with an API configuration that favors speed over cost efficiency, with identical quality, according to Anthropic.

Is fast mode included in my Max plan?

No. On every subscription plan, fast mode draws from usage credits, even when your plan has usage left. Without usage credits turned on, /fast doesn't work.

What happens to fast mode on Opus 4.8 on November 30, 2026?

According to Anthropic's notice to Team and Enterprise admins dated October 30, 2026, requests with fast mode on will fail after that date. Opus 4.8 at standard speed stays available. Either run /fast off or switch to Opus 5.5 with /model.

How much faster is fast mode?

Up to 2.5x, according to Anthropic. That is the vendor's figure, not a measurement of mine.

Where to go from here

Run /status and /fast to check how you are signed in and whether fast mode is on. On Opus 4.8, switch before November 30. And turn it on at the start of a session. If you are just getting started with Claude Code, read Claude Code: A Guide for Beginners. How others keep their Claude Code costs under control is what we discuss in my community AI Employee Framework.