Anthropic releases Claude Opus 5.5: first Claude 5.5 model; Fable 5.1-level on most work, ~40% cheaper than Opus 5

Why it matters

【Official release】

• Anthropic released Claude Opus 5.5 on 2026-09-22 — the first model in the Claude 5.5 family.
• Canonical page: https://www.anthropic.com/claude-opus-5-5 (also listed on the Newsroom).
• CMS featuredGrid/siteSettings _updatedAt≈2026-09-22T16:27:50Z is a verifiable approximate go-live; the page eyebrow is day-level only (September 22, 2026).
• API id: `claude-opus-5-5`; Anthropic says it is available on the Claude Platform and on Amazon Web Services, Google Cloud, and Microsoft Azure.

【Verifiable points】

• Official claims: performs at Claude Fable 5.1 level on most work; ~40% lower cost than Opus 5 on typical workloads.
• Pricing vs Opus 5: input $4 / output $20 per MTok (~20% lower each); cache reads $0.20 per MTok (~60% below Opus 5’s $0.50); cache writes $5.
• Strengths highlighted in agentic coding, computer use, and knowledge work.
• Pre-release external evaluation included Frontier Design and METR; Anthropic says Opus 5.5 is its strongest model to date on the automated behavioral audit.
• Because biology/cyber capabilities are comparable to Mythos/Fable-class models, it ships with safeguards similar to Fable 5.1 (cyber tasks often fall back to Opus 4.8, biology via Life Sciences Verification, plus distillation/preserved-thinking controls).
• Sonnet 5.5 and Haiku 5.5 are said to follow in the coming weeks.

【Our take】

No on-site Opus 5.5 event found; this is a new item, not a rehash. Facts limited to official visible text; benchmark tables and System Card details are not restated line-by-line here. do not touch e/663 or e/664.