Anthropic takes Opus 5.5 fast mode out of research preview: no waitlist or beta header, up to 2.5x faster output at $8/$40, twice standard price

Why it matters

【Key points】

Anthropic's October 15 release notes take fast mode for Claude Opus 5.5 out of research preview. No access request or waitlist; requests no longer need the fast-mode-2026-02-01 beta header (sending it still works); every usage tier gets a default fast-mode rate limit, separate from standard Opus limits.

Fast mode runs the same model and weights on a faster inference setup: up to 2.5x output tokens per second, with no change to time to first token. Set speed: "fast". Price is $8/$40 per million input/output tokens, twice Opus 5.5's standard $4/$20. Fast mode for Opus 5 and Opus 4.8 stays in research preview at $10/$50.

@ClaudeDevs added that Opus 5.5 fast mode is also on Microsoft Foundry (hosted on Azure), Google Cloud coming soon; fast mode in Claude Code is unchanged and uses extra usage credits.

【Verifiable】

Docs: https://platform.claude.com/docs/en/build-with-claude/fast-mode

Release notes (Oct 15): https://platform.claude.com/docs/en/release-notes/overview

Pricing: https://platform.claude.com/docs/en/about-claude/pricing

Announcement: https://x.com/ClaudeDevs/status/2110825294655562048

At publication the docs still say fast mode is Claude API-only and not on Microsoft Foundry, which conflicts with the post. Opus 5.5 launch: e/666.

【Our read】

1. A pricing and availability change, not a new model. Teams building latency-sensitive agents no longer wait on a list, and the separate limit doesn't eat standard quota.

2. The 2x premium is below Opus 5/4.8 fast pricing ($10/$50), nudging users to Opus 5.5; Opus 4.7 rejects fast-mode requests outright.

3. Only output speed improves, not first-token latency; long-input, short-output workloads gain little, so test on your own traffic first.