Skip to content

Claude fast mode: Opus 5.5 up to 2.5x faster, no waitlist

Claude fast mode for Opus 5.5 is out of research preview: up to 2.5x faster output, $8/$40 per million tokens, no waitlist. Here is how to turn it on.

Piotr Sobolewski4 min read

This article in other languages:العربيةDeutschEspañolFrançaisItaliano日本語Polski简体中文

In this article
  1. What changed?
  2. How does Claude fast mode work?
  3. How much does fast mode cost on Opus 5.5?
  4. How do we turn on fast mode?
  5. What to watch out for?
  6. FAQ

On October 15, 2026, Anthropic dropped the research preview label from Claude fast mode for Opus 5.5. No more waitlist, no more asking for access: just add speed: "fast" to the request, and Claude Opus 5.5 generates its answer up to 2.5 times faster, according to Anthropic. Same model, same weights, just a higher price per token.

What changed?

Fast mode for Opus 5.5 arrived with the model’s launch on September 22, but only as a research preview you had to get access to. Starting October 15, according to the Claude Release Notes:

  1. Opus 5.5 requests no longer need the beta header fast-mode-2026-02-01 (old requests that send the header keep working).
  2. Organizations on the Start, Build and Scale tiers get a fast mode rate limit right away. Accounts on the Evaluation tier get one once they move up a tier.
  3. Fast mode for Opus 5.5 also works in Claude in Microsoft Foundry, for Azure-hosted deployments.

Opus 5 and Opus 4.8 still have fast mode in research preview, so the waitlist and the beta header still apply there.

How does Claude fast mode work?

This is not a separate, “slimmed-down” model. According to the documentation, Anthropic runs the same Opus 5.5, just on a faster inference configuration, so its intelligence and capabilities stay the same.

Interestingly, the speedup applies to output tokens per second, not to time to first token. In other words, the answer doesn’t start any sooner, but once it gets going, it flies. Streaming shows this best.

How much does fast mode cost on Opus 5.5?

Pricing from the fast mode documentation:

Model Input (per million tokens) Output (per million tokens)
Claude Opus 5.5, fast mode $8 $40
Claude Opus 5.5, standard speed $4 $20

The standard-speed rates come from the Opus 5.5 announcement of September 22. So fast mode costs exactly twice as much, across the whole context window (including above 200K input tokens). Prompt caching and data residency multipliers stack on top of that price.

How do we turn on fast mode?

In the API, one parameter is all it takes:

response = client.messages.create(
    model="claude-opus-5-5",
    max_tokens=4096,
    speed="fast",
    messages=[{"role": "user", "content": "Refactor this module"}],
)
print(response.usage.speed)  # "fast" or "standard"

The usage.speed field in the response tells us which speed we actually got. Fast mode has its own rate limit: once we exceed it, the API returns a 429 error with a retry-after header, and the documentation shows how to fall back to standard speed in that case.

In Claude Code, we turn fast mode on with the /fast command (when it’s active, a ↯ icon shows up next to the prompt). Since v2.1.280, the default model for fast mode is Opus 5.5. On the Pro, Max, Team and Enterprise plans, fast mode is billed only through usage credits, outside the subscription limits (on Team and Enterprise, an organization Owner has to enable it first).

What to watch out for?

  1. Prompt cache. Switching between speeds invalidates the cache, because requests at different speeds don’t share cached prefixes. In Claude Code, turning fast mode on for the first time mid-conversation means paying the full rate for the entire context so far, which is why the documentation recommends enabling it at the start of a session.
  2. Batch API and Priority Tier. Fast mode doesn’t work with them.
  3. Clouds. Fast mode isn’t available on Claude Platform on AWS or in Anthropic-hosted Foundry deployments.

What is it best for? Anthropic points to interactive work: fast iterations on code and live debugging. For long autonomous tasks and CI pipelines, standard speed comes out cheaper. If we happen to be moving an app to new models (for example after the GPT-4o API deprecation), it’s worth checking right away whether response speed is what really matters somewhere in the product.

FAQ

Does Claude fast mode change answer quality?

No. According to Anthropic, it’s the same Opus 5.5 model with the same weights, just on a faster inference configuration.

How much does Claude fast mode cost?

For Opus 5.5: $8 per million input tokens and $40 per million output tokens, twice the standard speed ($4/$20).

How do I turn on fast mode in Claude Code?

With the /fast command or the "fastMode": true setting in the settings file. On subscription plans, usage credits need to be enabled.

Is fast mode available for Opus 5 and Opus 4.8?

Yes, but still in research preview: you have to request access or join the waitlist, and send the beta header.