Anthropic’s Claude Platform release notes for 15 October 2026 say fast mode is out of research preview for Claude Opus 5.5, and that you no longer need to request access. This is a runtime option on the existing model, not a new Opus release.
Requests to Claude Opus 5.5 no longer require 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 the Start, Build and Scale usage tiers. Fast mode for Claude Opus 5 and Claude Opus 4.8 remains in research preview.
The same notes say fast mode for Claude Opus 5.5 is available on Claude in Microsoft Foundry for deployments hosted on Azure.
What fast mode is (and is not)
Anthropic’s fast mode docs describe the feature as the same model weights with a faster inference configuration: “There is no change to intelligence or capabilities.” Claimed benefits focus on output tokens per second (OTPS) — “up to 2.5x higher” versus standard — not time to first token (TTFT). Opt in with speed: "fast" on the Messages request.
Do not describe this as a new model.
Pricing and limits
For Claude Opus 5.5, docs list fast mode at $8 / $40 per MTok (input / output), versus standard $4 / $20. Fast mode pricing stacks with prompt-caching and data-residency multipliers. Fast mode has a dedicated rate limit separate from standard Opus limits; exceeding it returns HTTP 429 with retry guidance.
Anthropic says fast mode is not available with the Batch API, Priority Tier, or currently on Claude Platform on AWS. Switching between fast and standard invalidates the prompt cache — the two speeds do not share cached prefixes.
Who can use it now
Because Opus 5.5 fast mode left research preview, Start/Build/Scale customers can call it without the waitlist path that still applies to Opus 5 and Opus 4.8 fast mode (those still need the beta header and access). Evaluation-tier orgs, the docs say, get Opus 5.5 fast mode once they move to Start, Build or Scale.
Checking which speed ran
The Messages response usage object includes a speed field reporting "fast" or "standard". Anthropic’s docs say that requesting speed: "fast" on an unsupported model returns an error, as does exceeding fast-mode capacity (429 or 529). Organisations can fall back to standard speed on rate limits, but that fallback misses the prompt cache.
Supported surfaces for Opus 5.5 fast mode include the Claude API (including Claude Managed Agents) and Claude in Microsoft Foundry on Azure-hosted deployments. Fast mode is not available for Foundry deployments hosted on Anthropic, according to the same docs table.
Context from earlier TSN coverage
Anthropic has been shipping both agent products and model-runtime controls. TSN covered Claude Code Review going generally available and, earlier, Claude Haiku 5.5. Fast mode GA for Opus 5.5 is the latency/cost dial for teams already standardised on that frontier model.
Why it matters
Agentic coding loops are often gated by how fast tokens stream once the model starts writing. A generally available fast path at 2× list token prices is a clear commercial trade: pay more per token to keep OTPS up, without migrating model IDs — provided cache behaviour and platform coverage (no Batch, no AWS Platform yet) fit the workload.
What’s still unknown
- When Opus 5 / 4.8 fast mode leave research preview
- Whether Claude Platform on AWS will gain fast mode
- Real-world OTPS versus the “up to 2.5x” claim under production tool-use loads
What to watch
Watch release notes for Batch/Priority/AWS Platform support and any change to default fast-mode rate limits by tier.
What this means
- Opus 5.5 fast mode is out of research preview — no access request; beta header optional.
- Same weights; up to 2.5× OTPS (Anthropic); priced at $8 / $40 per MTok.
- Opus 5 and 4.8 fast mode stay in research preview.
Sources
Related stories
- Anthropic Makes Claude Code Review Generally Available for Team and Enterprise
- Claude Haiku 5.5: Anthropic’s Small Model Gets Much Cheaper and Much Better at Using Computers
Follow @tsnmediahub on X for breaking AI news as it happens.
Follow on X
