Claude Code Fast Mode: How to Get 2.5x Faster Responses

Fast mode is a high-speed configuration for Claude Opus that prioritizes token generation speed over cost. It's not a different model and not a version with reduced reasoning: it's the same Opus, same quality, up to 2.5x faster. The tradeoff is a significantly higher cost per token.

It runs on Opus 5.5, Opus 5 and Opus 4.8, and since v2.1.280 the default /fast model is Opus 5.5. It is not available on Sonnet, Haiku or any other model. Fast mode on Opus 4.7 was removed on July 24, 2026, so switching to that model turns fast mode off automatically.

Personal note: on the medium-sized tasks I've tested it with, the speed difference is noticeable. What makes me hesitate isn't the price itself, it's where the price comes from: fast mode is always billed against your usage credits, even when you still have plan usage left. It isn't spending what you already pay for sooner, it's a separate bill from the first token.

Don't confuse fast mode with effort level adjustment. Lowering effort level does reduce reasoning quality in exchange for speed. Fast mode sacrifices nothing in quality, only cost.

Result:

> /fast

Fast mode ON

A small ↯ icon sits next to the prompt while fast mode is active.

One detail catches people out: if you were on another model, turning fast mode on switches you to Opus automatically. It works both ways, since switching to a model that doesn't support it turns fast mode off by itself.

Setup

1. Enable fast mode

Type /fast and press Tab to toggle it on or off:

/fast

The VS Code extension has no such command: it follows your fastMode setting and offers a Toggle fast mode command when the selected model supports fast mode.

2. Disable when you don't need it

The same command toggles it off. When disabled you stay on Opus: it doesn't revert to your previous model. To switch models, use /model.

3. Enable by default (optional)

To keep it always on, add to your user settings:

{
  "fastMode": true
}

4. Or the opposite, start every session with it off

Useful if you run many concurrent sessions and don't want fast mode leaking into all of them:

{
  "fastModePerSessionOptIn": true
}

Your preference is still saved, so removing the setting restores the persistent behavior.

Turn it on at the start, not mid-conversation

The first time you enable fast mode in a conversation, you pay the full uncached input price, at the fast mode rate, for the entire context accumulated up to that point. The deeper into the conversation you are, the more that first moment costs. The charge applies once per conversation, so toggling it off and back on later does not repeat it.

That is a cache miss in the full sense, and since v2.1.251 you do not have to take my word for it: /cost files it under likely cause: fast mode toggled, with the tokens it took to rebuild the prefix.

Requirements

You need all three at once:

• A subscription (Pro/Max/Team/Enterprise) or the Anthropic Console. It does not work on Amazon Bedrock, Google Cloud Agent Platform, Microsoft Foundry or Claude Platform on AWS.
• Usage credits turned on for your account, which is what allows billing beyond what your plan includes. On individual accounts you turn it on in the Console billing settings.
• Owner enablement on Team and Enterprise. It ships disabled for the organization, and until an Owner enables it, /fast answers Fast mode has been disabled by your organization.

Reference

When to use it

• Rapid iteration: code changes where waiting 30 seconds matters.
• Live debugging: interactive sessions where every second counts.
• Deadline-driven work: when speed justifies the extra cost.

For long autonomous tasks, batch processing, or CI/CD pipelines, standard mode is more cost-efficient.

Official docs: Speed up responses with fast mode