Anthropic releases Claude Opus 5.5

2026-09-22 — Tuesday

Anthropic today released Claude Opus 5.5, the first model of the Claude 5.5 generation. The pitch is simple: performance at roughly the level of Claude Fable 5.1 on most work, at a price about 40% below Opus 5 on typical workloads, with output generated more than 30% faster. Sonnet 5.5 and Haiku 5.5 are due in the coming weeks.

Benchmarks

The gains are concentrated in agentic coding, knowledge work and computer use. On several agentic benchmarks Opus 5.5 edges past Fable 5.1, the more expensive top tier:

Benchmark	Opus 5	Fable 5.1	Opus 5.5
Terminal-Bench 4.0	52.3%	55.8%	66.4%
FrontierCode v1.1 Main	48.0%	50.3%	54.4%
CursorBench 4.0	—	51.8%	57.8%
GDPval-AA v2.1 (Elo)	1708	1735	1846

Anthropic itself cautions that at this capability level “benchmark margins have become a less reliable guide to real-world differences” — worth keeping in mind before reading too much into a few points either way.

Pricing

$4 per million input tokens, $20 per million output — down 20% from Opus 5’s $5 / $25.
Cache reads: $0.20 per million — 60% cheaper than Opus 5; cache writes $5 per million.
Fast mode: $8 / $40 per million, for roughly 2.5× faster output.
Batch API: the usual 50% discount.

As with Fable 5.1 three weeks ago, the cache discount is the part that matters most for long-running agents, where re-reading the same context on every turn dominates the bill.

Specs and features

Model ID: claude-opus-5-5
Context window: 1M tokens
Max output: 128K tokens synchronous; 300K via the Batches API (beta)
Adaptive thinking: always on, tuned with the effort parameter
Knowledge cutoff: June 2026
Writing: noticeably clearer and more direct than Opus 5

Anthropic’s long-horizon coding examples include a 680,000-line migration completed in under a day, and an audit of a 200,000-line codebase in under three hours (versus 20+ hours for Opus 5).

Safety

Highest scores to date on Anthropic’s automated behavioral audit.
85% fewer attempts to circumvent containment boundaries than Opus 5.
Cybersecurity and biology safeguards matching Fable 5.1, with verified-access programs for security and life-sciences professionals.

Availability

Opus 5.5 is live today on the Claude apps, the Claude API, Amazon Bedrock and Claude Platform on AWS, Google Cloud, and Microsoft Foundry, with expanded usage limits on Pro, Max, Team and Enterprise plans.