Opus 5.5 ships: 40% cheaper, 66.4% on Terminal-Bench, API at $4/$20
mark_k · x · 2026-09-23
Mark Kretschmann rounds up Anthropic's official numbers for Claude Opus 5.5:
- 40% lower costs on typical workloads than Opus 5
- Over 30% faster output
- 66.4% on Terminal-Bench 4.0, versus Opus 5's 52.3% and GPT-6 Astra's 57.9%
- Clearer writing and stronger performance on long, complex coding tasks
API pricing drops to $4/$20 per million input/output tokens. Pro, Max and Team users also get higher five-hour limits plus a rate-limit reset they can save for later.
Related event: Anthropic Launches Claude Opus 5.5: Stronger, 40% Cheaper, Tops AA Index(259 posts)→
More from Models
- Claude's connector management criticized: Chat-only visibility, buried in 'Customize' tab — tristanbob · 2026-11-09
- Devs: Sora makes indistinguishable video while image gen still looks like slop — Angaisb_ · 2026-11-09
- Rumor: NDA on rumored fable 5.5 model reportedly ending, launch may land next week — iruletheworldmo · 2026-11-09
- GameFeel Bench v1.4: Qwen-4-Flash-Early leads at 49.9 after score fix — Sea_Emergency_8458 · 2026-11-09
- Google AI Studio shares new tutorial and more info — DuRuofei · 2026-11-09
- Indian "built in India" Alpie 9B exposed as a near bit-for-bit copy of Qwen3.5-9B — dotey · 2026-11-09