The Lab Notes

Anthropic

Claude Opus 5.5 fast mode no longer needs an access request

Anthropic said Oct. 15, 2026, that fast mode for Claude Opus 5.5 is out of research preview, with no beta header and a default rate limit in every tier.

The Lab Notes ·

A large chrome stopwatch with a swirl of pink, orange and purple blur on its face sits at the center of a background split down the middle, black and navy blue with thin white contour lines on the left and a repeating pattern of small red, yellow, green and blue stopwatches on black on the right.

Anthropic has taken fast mode for Claude Opus 5.5 out of research preview. Developers no longer have to request access, the company said in the Claude Platform release notes.

The entry is dated Oct. 15, 2026. Requests to Opus 5.5 no longer need the fast-mode-2026-02-01 beta header. Requests that still send it keep working. Opus 5.5 now has a default fast mode rate limit in every usage tier.

Fast Mode Costs Twice the Standard Price

Anthropic charges $8 per million input tokens and $40 per million output tokens for Opus 5.5 in fast mode, according to its fast mode docs as of Oct. 15, 2026. Standard Opus 5.5 pricing is $4 and $20, as The Lab Notes reported at launch. Fast mode costs twice as much on each. The multiplier applies across the full context window, including requests over 200,000 input tokens. Prompt caching and data residency multipliers apply on top.

Fast mode runs the same model with a faster inference configuration, producing up to 2.5x more output tokens per second than standard speed. The gain is in output speed, not time to first token. The weights and behavior are the same. Developers opt in by setting speed: “fast” on the request.

Opus 5 and Opus 4.8 Stay in Preview

Anthropic keeps fast mode for Claude Opus 5 and Claude Opus 4.8 in research preview. Requests to those models must keep sending the beta header. Fast mode on them costs $10 per million input tokens and $50 per million output tokens.

Fast mode is not available on Opus 4.7, where speed: “fast” requests return an error. On Opus 4.6, requests with the beta header run at standard speed and are billed at standard rates. Requests without it return a 400 error.

Fast mode runs only on the Claude API, including Claude Managed Agents. Amazon Bedrock, Claude Platform on AWS, Google Cloud and Microsoft Foundry do not offer it.

Fast Mode Has Its Own Rate Limit

Fast mode has a dedicated rate limit, separate from the standard Opus limits. The API returns a 429 error with a retry-after header on requests that exceed it. Responses carry anthropic-fast-* headers that show the fast mode rate limit status. The limits appear on the Rate limits page in the Claude Console.

Our View

We start from the separate rate limit. Requests sent with speed: “fast” count against their own allowance and leave the standard Opus allowance alone. Developers can send the calls where a person waits on the output, such as a chat reply, at fast speed. Work nobody waits on can stay at standard speed and standard price. The doubled rate lands only on the calls that go fast. Developers who were waiting on an access request can now test that split in every usage tier. They will need to watch for 429 errors on the fast calls and decide whether to retry or send the call at standard speed. The split works only if the default limit in a developer’s tier is large enough for the traffic they want to send fast.