Does Claude Opus 5.5 Match Fable 5.1 at 40% Lower Cost?
VerifiedAnthropic released Claude Opus 5.5 on September 22, 2026, pricing it at $4/$20 per million input/output tokens — 40% cheaper than Opus 5's $5/$25 — while positioning it as performing on par with the larger Claude Fable 5.1 on most tasks. Anthropic's own benchmark data shows Opus 5.5 scoring 81.8% on OSWorld 2.0 versus Fable 5.1's 80.7% and Opus 5's 74.0%, and beating Fable on Terminal-Bench 4.0 (66.4% vs 55.8%). TechCrunch independently corroborated the pricing cut. AI Ledger rates this Verified.
Frequently asked questions
Has Anthropic confirmed these pricing and benchmark figures itself?
Yes. The figures come directly from Anthropic's own product page announcing Claude Opus 5.5 on September 22, 2026, making this a primary-source confirmation rather than secondhand reporting.
How much cheaper is Claude Opus 5.5 than Opus 5?
Anthropic priced Opus 5.5 at $4 per million input tokens and $20 per million output tokens, down from Opus 5's $5 and $25 — a 40% reduction on typical workloads, with cache-read pricing also dropping from $0.50 to $0.20 per million tokens.
Does Opus 5.5 actually beat Claude Fable 5.1, or just match it?
Results are mixed by benchmark. Anthropic's own data shows Opus 5.5 scoring higher than Fable 5.1 on OSWorld 2.0, Terminal-Bench 4.0, and GDPval-AA v2.1 — though Anthropic frames the overall comparison as performing 'at the level of' Fable rather than uniformly beating it, since Fable remains the larger model overall.
Were these benchmark scores independently verified?
Not yet by a third party. TechCrunch corroborated the pricing cut and Anthropic's overall characterization of the release, but reported the specific benchmark numbers as Anthropic's own self-published results. AI Ledger will update this page if independent evaluators publish different figures.