Anthropic launches Claude Opus 5.5, priced 40% below Opus 5 and claimed to match Fable 5.1 on most work
2026-09-22
Anthropic released Claude Opus 5.5 (claude-opus-5-5) on Sept 22, 2026: a 1M-token context window by default, 128k max output tokens, and always-on adaptive thinking, priced at $4/$20 per MTok versus Opus 5's $5/$25. Anthropic states Opus 5.5 "performs at the level of Claude Fable 5.1 on most work" while costing 40% less to run than Opus 5. It is available via the Claude API, Amazon Bedrock, Claude on Google Cloud, and Microsoft Foundry.
Significance 4: A new flagship-tier release with a substantial cost cut and a lab-reported quality claim matching Anthropic's top model moves tracked cost/reasoning/coding axes; capped at 4 since the quality claim is lab-reported and not yet independently reproduced.
Implications · machine-drafted, not owner judgment
If Opus 5.5's quality claim holds up independently, it directly challenges the current coding-agent and daily-driver verdicts (both held by Claude Opus 4.8) on economics alone — a materially cheaper model claimed to perform near the top of Anthropic's own lineup changes the cost side of the verdict calculus before any capability delta is confirmed. It also gives the owner a meaningfully cheaper near-frontier option for this tracker's own pipeline calls and other AI-engineering work.
- Independent SWE-bench Verified and Terminal-Bench scores for Opus 5.5 against Opus 4.8
- Real-world agentic recovery-from-bad-state behavior, the deciding factor in the current coding-agent verdict
- Whether Opus 5.5 or Opus 4.8 becomes the default in Claude Code
Sources
- primary Introducing Claude Opus 5.5 retrieved 2026-09-25
- primary Claude Platform release notes — September 22, 2026 retrieved 2026-09-25
- curator [AINews] The Future of Latent Space retrieved 2026-09-25
Models
Affected axes
Cost · Coding · Agentic / tool use · Hard reasoning