Fast mode for Claude Opus 5.5 is now one parameter away: set speed to fast. It costs $8 in and $40 out per million tokens, speeds up output rather than the first token, and every switch between fast and standard speed is a prompt cache miss. Route by workflow, not by request.
Anthropic's documentation now says fast mode for Claude Opus 5.5 needs no beta header and no access request. You opt in by setting speed to fast on a Messages API request. Requests that still send the old fast-mode-2026-02-01 header keep working, so existing code will not break, but new code can drop it.
Know what you are buying. Fast mode for Opus 5.5 costs $8 per million input tokens and $40 per million output tokens, and Anthropic says it delivers up to 2.5x higher output tokens per second than standard speed. The documentation is explicit that the gain is in output speed, not time to first token. A short answer will not feel faster. A long code diff, a drafted document, or a verbose agent step will.
The trap is caching. Anthropic's documentation says switching between fast and standard speed invalidates the prompt cache, because requests at different speeds do not share cached prefixes. A workflow that flips speed per request, or one that falls back to standard speed whenever fast capacity runs out, pays full input price on every switch. If you lean on a large cached system prompt, that can erase the savings you were counting on.
Fast mode also has its own rate limit, separate from standard Opus limits. When you exceed it, the API returns a 429 with a retry after header, and the SDK retries up to 2 times by default. If you would rather fall back than wait, set max_retries to 0 on the fast request, catch the rate limit error, and resend without the speed parameter, accepting the cache miss that comes with it. Fast mode is not available with the Batch API or on Claude Platform on AWS.
Finally, check what you were billed for. The response usage object carries a speed field that reads fast or standard, which is the cleanest way to confirm a request actually ran fast before you attribute a latency win to it. If you still run Opus 5 or Opus 4.8 in fast mode, those remain in research preview at $10 in and $50 out per million tokens, so migrating to Opus 5.5 is cheaper as well as simpler.
Try it today
- Pick the workflows where users wait on long Opus 5.5 output, such as code generation, drafting, and long agent steps, and enable fast mode only there by setting speed to fast.
- Keep each workflow on one speed. Do not toggle per request, because every switch between fast and standard speed is a prompt cache miss.
- Decide your rate limit policy: let the SDK retry (2 times by default) or set max_retries to 0 and fall back to standard speed, budgeting for the cache miss.
- Log usage.speed on every response and compare cost and completion time against standard speed before rolling fast mode out widely.
- Move any remaining Opus 5 or Opus 4.8 fast mode traffic to Opus 5.5 to drop the beta header and the higher $10 and $50 rates.