In brief
- On October 15, 2026, Anthropic moved Claude Opus 5.5 fast mode out of research preview.
- Up to 2.5x output tokens per second, according to the company, at $8 and $40 per million tokens.
- No access request is needed in the API. Available on the Claude API and Microsoft Foundry on Azure.
Anthropic moved Claude Opus 5.5 fast mode out of research preview, so it is now available without an access request.
AI Desk · Published October 16, 2026
First look: this report is based on Anthropic’s official documentation and release notes and has not yet been tested hands-on by Maitz.
The change appears in the Claude Platform release notes dated October 15, 2026. Claude Opus 5.5 itself launched on September 22, when fast mode was available only as a research preview.
What changed with Claude Opus 5.5 fast mode?
Set speed: "fast" on a request to claude-opus-5-5 and, according to Anthropic, you get up to 2.5x higher output tokens per second. The model, weights and behavior are the same. It is a different configuration of the same model.
The speedup targets output rate, not time to first token. It shows most in streaming. Claude Opus 5.5 now has a default fast mode rate limit on the Start, Build and Scale tiers.
| Item | According to Anthropic |
|---|---|
| Standard price, Opus 5.5 | $4 input / $20 output per million tokens |
| Fast price, Opus 5.5 | $8 input / $40 output per million tokens |
| Fast price, Opus 5 and 4.8 | $10 input / $50 output, still in research preview |
| Availability | Claude API, including Managed Agents, and Microsoft Foundry on Azure |
| Not available | Batch API, Amazon Bedrock, Claude Platform on AWS, Google Cloud |
The premium applies across the full context window, including requests over 200K input tokens. It stacks with prompt caching and data residency multipliers.
How does it work in Claude Code?
According to the Claude Code docs, the /fast command toggles the mode. From version 2.1.280, Opus 5.5 is the default model for fast mode. On Team and Enterprise plans, an owner must enable it first.
On Pro, Max, Team and Enterprise plans, fast mode is billed from usage credits and is not included in the plan’s usage limits. For Console organizations, the docs say access must be provisioned.
When is paying double worth it?
Speed matters most in interactive work: rapid iteration and live debugging. For long background jobs or batches the premium is harder to justify, and batches do not support the mode at all.
Our take: switch it on only where a person is waiting for the output. Everywhere else, standard Opus 5.5 pricing is still lower than Opus 5.
What we don’t know yet
- Independent speed measurements on real workloads. “Up to 2.5x” is the company’s figure.
- When fast mode for Opus 5 and Opus 4.8 leaves research preview.
- Whether the mode will reach Bedrock and Google Cloud.
What it means for you
- Developers: add
speed: "fast"only to endpoints where a user is waiting, and measure cost against latency before expanding. - Claude Code users: try
/fastin an interactive session (billed from usage credits on subscription plans) and turn it off for long tasks. More in our Claude guide. - Teams on Bedrock or Google Cloud: the mode is not available there today. You would need the Claude API or Foundry.
- Prices: all in USD. For a smaller, cheaper model see Claude Haiku 5.5.
More on Maitz
Sources
- Anthropic: Claude Platform release notes, September 15 to October 15, 2026, platform.claude.com
- Anthropic: Fast mode, Claude API docs, platform.claude.com
- Anthropic: Pricing, Fast mode pricing, platform.claude.com
- Anthropic: Speed up responses with fast mode, Claude Code Docs, code.claude.com
- Verified: October 16, 2026
