What does Claude Code's /fast mode change about an Opus session, how is it billed, and when is lowering effort the better lever?
answer
- same model, different API configuration
- research preview, Opus only
- usage credits, not plan limits
- enable it at the start
- speed for money versus depth
basics
~20 sFast mode, a research preview, runs Opus with an API configuration tuned for speed, up to 2.5x faster at the same quality but a higher per-token price; subscribers pay from usage credits. Lowering effort instead buys speed by thinking less.
solid answer
~50 s`/fast` toggles fast mode, a research preview that serves the same Opus model through a speed-prioritizing API configuration: up to 2.5x faster, identical quality, higher price per token. It works on Opus 5.5, Opus 5 and Opus 4.8; turning it on from another model switches the session to Opus, and the `↯` icon marks it as active. On Pro, Max, Team and Enterprise it bills only to usage credits, never the plan's included limits, and Team and Enterprise need an Owner to enable it; Console organizations pay per token. It is not offered on Bedrock, Agent Platform, Foundry or Claude Platform on AWS. Turn it on at the start of a session, because the first fast turn pays the fast rate on the whole uncached context. Lowering effort is the other lever: cheaper and faster, but potentially lower quality on complex tasks.
go deeper
Know that /fast makes Opus answer faster for more money, that it is a research preview, and that the ↯ icon shows it is on.
Explain which models and providers support it, that subscribers pay from usage credits only, and why turning it on early in a session is cheaper.
Pick between fast mode and lower effort for a given workload, and explain organization-level blocks such as Owner enablement and per-session opt-in.
Set a team policy for a priced preview: where speed justifies the spend, whether to require per-session opt-in, and how to watch credit consumption.
## What fast mode is **Fast mode** is a **research preview**: its behavior, pricing and availability may change. It is not a different or smaller model. Claude Code sends requests to the same Claude Opus model with a different API configuration that prioritizes speed over cost efficiency, making responses up to 2.5x faster with identical quality and capabilities. The price per token is higher. ## Where it works | Question | Answer at Claude Code 2.1.286 | |---|---| | Which models | Opus 5.5 (the fast-mode default since v2.1.280), Opus 5, Opus 4.8 | | Which models not | Sonnet, Haiku and other models; Opus 4.7 lost it in July 2026 | | Which access paths | the Anthropic Console API and Claude subscription plans | | Not available on | Amazon Bedrock, Google Cloud's Agent Platform, Microsoft Foundry, Claude Platform on AWS | ## How it is billed - **Subscription plans** (Pro, Max, Team, Enterprise): fast mode is billed only to **usage credits** and is not part of the plan's included rate limits. It draws from credits even when plan usage remains, and `/fast` refuses until usage credits are turned on. - **Team and Enterprise**: fast mode is off for the organization until an **Owner** enables it. - **Claude Console**: the organization pays per token with the rest of its API usage, and needs fast-mode access provisioned. - **Turning it on late** costs more. The first fast request in a conversation reads the whole history uncached at fast-mode rates, so enabling it at the start of a session is cheaper than enabling it deep into one. The charge applies once per conversation; toggling off and on again later does not repeat it. Why that happens is a prompt-caching question. ## Turning it on and off 1. Run `/fast`, press Space to toggle, then Enter, or set `"fastMode": true` in user settings. 2. If the current model does not support fast mode, Claude Code switches to Opus; you see "Fast mode ON" and a `↯` icon by the prompt. 3. Turned on mid-turn, it takes effect from the next turn; the running turn finishes at its original speed. 4. Running `/fast` again turns it off but leaves you on Opus; use `/model` to move elsewhere. Other behaviors worth knowing: - By default the setting **persists across sessions**. `fastModePerSessionOptIn: true` makes every session start with it off. - Switching to a model without fast mode turns it off; switching back to a supported Opus turns it on again only when your saved preference is on. - `CLAUDE_CODE_DISABLE_FAST_MODE=1` disables it entirely, and managed settings can set `fastMode: false`. - Fast mode has its **own rate limits**, shared across the supported Opus models. Hitting them drops you to standard speed and pricing with a grey `↯` during the cooldown, after which it re-enables automatically. ## Fast mode or a lower effort level? | Lever | Effect | |---|---| | Fast mode | same model quality, lower latency, higher cost | | Lower effort | less thinking, faster responses, potentially lower quality on complex tasks | Fast mode fits interactive work where waiting is the bottleneck: rapid iteration, live debugging, deadline work. Standard speed fits long autonomous tasks, batch or CI pipelines and cost-sensitive work. The two combine: fast mode with a lower effort level gives the quickest turnaround on straightforward tasks. If the work is simple and money matters, lower effort is the cheaper lever; if quality must hold and someone is waiting, fast mode is the one that keeps the model's full reasoning. ## Seeing what it cost Where fast-mode spend shows up depends on how you signed in, which `/status` tells you: - **Pro and Max**: the usage-credits section of the claude.ai usage settings, which includes fast mode without breaking it out. - **Team and Enterprise**: `/usage` shows your own usage-credits spend; the organization's spend is visible to admins. - **Claude Console**: the Usage and Cost pages can group by speed to separate fast mode from standard usage. In non-interactive `claude -p` runs, `/fast` works only when the session was launched with fast mode in its `--settings` value, and then only for that session.
- A Team-plan developer runs /fast and sees that fast mode has been disabled by the organization. What are the likely causes?On Team and Enterprise fast mode is off until an Owner enables it in the admin settings. The same message appears when managed settings set `fastMode: false`, or when managed settings set `fastModePerSessionOptIn: true` and the request comes from somewhere other than an interactive terminal session. An `availableModels` list that excludes the fast-mode Opus model produces a different message about allowed models.
- What happens when you hit the fast mode rate limit mid-session?Claude Code falls back to standard speed and pricing automatically, the `↯` icon turns grey during the cooldown, and fast mode re-enables when the cooldown expires. You keep working without intervention, or run `/fast` to turn it off yourself. Fast mode limits are separate from standard Opus limits and shared across the supported Opus models.
Fast mode is paying for an express lane rather than asking the cook to skip steps: the same dish arrives sooner at a higher price, while lowering effort is the shortened recipe that is cheaper and quicker but may cut corners.
- Fast mode swaps in a smaller, lower-quality model to answer faster
- Fast mode usage counts against a Max plan's included limits
- Fast mode works on Sonnet for cheap speed-ups
- Turning fast mode on mid-conversation costs the same as at the start
- Fast mode is the same thing as setting effort to low
Assumes Claude Code 2.1.286: fast mode is a research preview on Opus 5.5 (its default since v2.1.280), Opus 5 and Opus 4.8.