What Claude Code Fast Mode Does, What It Costs, and When to Turn It On

When you use Claude Code to edit or debug software, Fast Mode makes Claude generate its response faster in exchange for higher token prices. A token is a small unit of text that Claude reads or writes. Fast Mode targets the speed at which Claude produces output; it does not guarantee that your entire coding task will finish faster.

Turn it on when you are actively waiting for Claude and want to keep an interactive session moving. Leave it off for long unattended work or when the extra cost outweighs the time saved. Anthropic advertises up to 2.5 times faster output generation, but that is a maximum output rate, not a promise that every task will take 2.5 times less time.

What Fast Mode changes

Fast Mode changes how a supported Claude model generates responses to prioritize speed. Anthropic says the mode keeps the same model, capabilities, and quality. The published documentation does not verify that claim with independent comparisons across coding tasks. Fast Mode speeds up output generation, but it does not guarantee a faster first token or a faster task from start to finish. Claude may still need to inspect files, run tools, or wait for other operations.

Fast Mode targets response speed and cost. Lowering Claude’s effort level changes how much reasoning Claude applies, which can affect answer quality on complex tasks. You can combine the settings, but reducing effort is not an equivalent substitute for Fast Mode. Anthropic describes these distinctions in its Claude Code Fast Mode documentation and Platform documentation.

Model support changes over time. As of November 11, 2026, Claude Code lists Claude Fable 5.2 and Claude Opus 5.5 as available for Fast Mode; Opus 5 and Opus 4.8 are listed as research preview. The same documentation says Opus 4.8 Fast Mode is scheduled for removal on November 30, 2026. Check the current model and provider requirements before relying on Fast Mode for a particular setup.

What Fast Mode costs

Anthropic charges for Fast Mode by input and output tokens, rather than by session or toggle. The listed rates below are U.S. dollars per million tokens (MTok). For the models shown, Fast Mode doubles both the input and output rates.

For example, one million input tokens on Opus 5.5 cost $4 at standard speed or $8 in Fast Mode; one million output tokens cost $20 or $40. Your total depends on how many tokens Claude reads and generates. Caching and other pricing modifiers also affect the bill, so the table does not predict a fixed cost per session. Anthropic’s pricing documentation describes the applicable rates and modifiers.

Subscription billing also matters. Pro, Max, and Ultra users pay for Fast Mode through usage credits, even if they have not used all the usage included in their subscription. Fast Mode spending counts toward the monthly amount shown in Settings → Usage. Team and Enterprise organizations also use usage credits, and an organization owner must enable Fast Mode. Claude Console users pay API rates through their organization; the free Evaluation plan does not support Fast Mode. Subscription users need usage credits enabled. Turning them on allows charges beyond the usage included in a plan.

Turning on Fast Mode partway through a conversation can add cost. Claude Code says the first switch to Fast Mode bills the context from the existing conversation at the Fast Mode uncached input rate. Fast Mode and standard speed use separate caches, so switching between them can cause Claude to read context without using a cache again. If you know you need Fast Mode, turn it on near the start of the session.

When to turn it on

Fast Mode makes the most sense when you are actively working with Claude and waiting for its streamed response. For example, turn it on while debugging a live issue and making several quick code changes, or during a short, time-sensitive editing session. The value comes from getting Claude’s output sooner so you can respond sooner. The overall task still depends on what Claude needs to do after generating that output.

Standard speed is generally the better choice for long tasks that run on their own. The same applies to batch work and CI/CD jobs, as well as requests where cost matters. These tasks get less benefit from faster responses when no one is waiting on each one. Fast Mode is unavailable through the Batch API.

To toggle the setting in the Claude Code CLI, run:

/fast

Press Space to turn it on or off, then press Enter to confirm. You can run /fast again or use /fast off to turn it off. Claude Code confirms activation with Fast mode ON and a ↯ icon beside the prompt. The VS Code extension also provides a Toggle fast mode command when the selected model supports the feature.

If the toggle is unavailable, check whether your selected model and provider support Fast Mode, whether usage credits are enabled, and whether your account or organization permits it. An organization’s settings can block access even when the model supports the feature.

What I'm building

Delegate tasks. Get software.

Give Vroni a GitHub issue, bug report, spec, or rough idea. It reads the repo, plans the change, writes code, runs checks, and works toward a review-ready pull request.

Take a look at vroni.com