On 15 October 2026, the Claude Platform release notes say fast mode is out of research preview for Claude Opus 5.5, and that you no longer need to request access. Requests to claude-opus-5-5 no longer require the fast-mode-2026-02-01 beta header. Requests that still send the header keep working. The same note says fast mode for Claude Opus 5 and Claude Opus 4.8 remains in research preview, and those requests still need the header.
The ClaudeDevs post at 20:09 UTC the same day says the waitlist is gone, that this is the same Opus 5.5 at up to 2.5 times the speed, at $8 and $40 per million tokens, and that Google Cloud is coming soon. An earlier account of Opus 5.5 described fast mode as a paid research preview on the Claude API.
Same weights, higher output rate
The fast mode docs say fast mode runs the same model with a faster inference configuration, with the same weights and behavior. That sameness is Anthropic’s description of the setting. You opt in with speed: "fast" on claude-opus-5-5. Opus 5.5 does not need the beta header or an access request.
Anthropic says the setting delivers up to 2.5 times higher output tokens per second than standard speed. The docs say the gain is in output tokens per second, and that time to first token is not where the benefit sits. The 2.5 times figure is Anthropic’s. This account did not time a request.
The listed price is twice standard
On the pricing page, standard Claude Opus 5.5 is $4 per million input tokens and $20 per million output tokens. Fast mode for Opus 5.5 is $8 and $40. Those fast rates are twice the standard input and output rates on that page. The fast-mode page prints the same $8 and $40. Opus 5 and Opus 4.8 fast mode, still in research preview, are listed at $10 and $50.
Fast pricing applies across the full context window, including requests over 200,000 input tokens. Prompt-caching and data-residency multipliers stack on top of the fast rates. A cache read on standard Opus 5.5 is $0.20 per million tokens on the pricing page, which that page calls 5 percent of the standard input price.
Switching between fast and standard invalidates the prompt cache. The docs say requests at different speeds do not share cached prefixes.
Where a fast request is accepted
The fast-mode table lists Opus 5.5 on the Claude API, including Claude Managed Agents, and on Claude in Microsoft Foundry for deployments hosted on Azure. It says fast mode is not available for Foundry deployments hosted on Anthropic. The 15 October release notes repeat the Azure-hosted Foundry line.
The pricing page’s cloud sentence is slightly different. It says fast mode is on the first-party Claude API and, for Opus 5.5, on Azure-hosted Foundry, and that it is not available on Claude Platform on AWS or on partner-operated cloud platforms. The fast-mode page also excludes the Batch API, a Priority Tier commitment, and Claude Platform on AWS. Google Cloud is absent from both tables. The “coming soon” line is the ClaudeDevs post. It is not a sentence in the docs opened on 16 October.
Opus 5.5 has a default fast rate limit on the Start, Build, and Scale usage tiers, separate from standard Opus limits. Organizations on the Evaluation tier can use it once they move to Start, Build, or Scale. When the fast limit is exceeded, the API returns a rate-limit error. The docs name separate fast input and output token headers on the response.
Claude Code uses a different switch
Fast mode in Claude Code is the /fast command. That page still calls fast mode a research preview whose pricing and availability may change. The ClaudeDevs post says fast mode in Claude Code remains unchanged and uses extra usage credits. The Code page says subscription plans (Pro, Max, Team, and Enterprise) pay for /fast from usage credits, outside the subscription’s included limits, and that Team and Enterprise need an Owner to enable it. It also says this Claude Code path is not available on Amazon Bedrock, Google Cloud’s Agent Platform, Microsoft Foundry, or Claude Platform on AWS. That sentence is about /fast in Claude Code. API availability on Azure-hosted Foundry is the separate path above.
The Campfire
No commentsNobody has pulled up a log by this one yet. Be the first to say what you make of it.
Held for the desk. It appears after a look.