Anthropic removes fast mode for Claude Opus 4.6, silently downgrades to standard speed

Anthropic quietly completed the removal of fast mode support for Claude Opus 4.6 on June 29, 2026, following through on a deprecation notice it issued when Claude Opus 4.8 launched last month. The change is notable for its soft landing: rather than throwing an error, the API now silently runs requests at standard speed.

What's new

As of June 29, any API request to claude-opus-4-6 that includes speed: "fast" will no longer execute at fast speed or attract premium pricing. Instead:

• The request runs at standard speed and is billed at standard rates
• No error is returned — the call succeeds silently
• The response's usage.speed field reports the actual speed used

Developers who want to continue using fast mode need to migrate to Claude Opus 4.8, where fast mode is available as a research preview.

Context

Anthropics fast mode for Opus 4.6 launched as a research preview in early 2026, offering significantly faster output token generation at premium pricing for customers who needed lower latency on heavy workloads. When Opus 4.8 shipped on May 28, Anthropic simultaneously deprecated fast mode for Opus 4.6 and set removal for approximately 30 days later.

This removal is handled differently from the upcoming Opus 4.7 fast mode removal. Claude Opus 4.7's fast mode was deprecated on June 25 with a hard July 24 cutoff — after that date, requests with speed: "fast" on Opus 4.7 will return an error. The Opus 4.6 removal instead chose a silent downgrade path, reducing disruption for teams that hadn't yet migrated.

Anthropics model deprecation pattern is consistent: announce removal alongside a successor launch, give roughly 30 days, then enforce. Claude Sonnet 4 and Claude Opus 4 were retired on June 15 after a two-month notice period.

Why it matters

For developers still running production workloads on Opus 4.6 with fast mode enabled, this change means their requests are now slower than expected and billed at standard rates — without any error alerting them to the change. The silent downgrade is easier to miss than an outright error.

Teams that rely on fast mode latency for user-facing features should audit their model configuration immediately. Migrating to Claude Opus 4.8 restores fast mode access; Opus 4.8 also brings a 1M token context window by default, improved performance, and lower minimum prompt caching thresholds.