Claude Opus 5.5 tops Artificial Analysis Intelligence Index at 58, with 20% price cut to $4/$20
geoffwolfe · x · 2026-09-26
Artificial Analysis reports Claude Opus 5.5 takes the top spot on its Intelligence Index with a max-effort score of 58, several points above the prior best, while Anthropic cut pricing to $4/$20 per 1M tokens (from $5/$25) and cache reads from $0.50 to $0.20.
- Leads on six of ten evaluations: Humanity's Last Exam 61.4% (prior best 59.1%), SciCode 66.9% (prior best 63.1%); Terminal-Bench 4.0 at 59.6%, level with GPT-6 Astra (xhigh) and +11 over Opus 5
- Agentic knowledge work lead: Elo 1822 on private eval AA-Briefcase, +143 over Fable 5.1, first time Anthropic surpasses GPT-5.6 Sol on presentation quality
- Cost: 119k output tokens per task vs 73k for Opus 5 and 27k for GPT-6 Astra, yet cost per task stays flat
- Four of five effort levels sit on the intelligence-vs-cost frontier; still behind on CritPt, AA-LCR, GDP.pdf
More from Models
- Qwen Team Details Qwen3.8-Omni-Flash: an Omni-Modal Agent Model with 1M Context — dair_ai · 2026-09-26
- GPT-6 Sol hits #6 on Agent Arena with +7.7% net improvement at 56% less cost per task — arena · 2026-09-26
- Matt Shumer says Opus 5.5 spontaneously added helicopter easter eggs to his website — mattshumer_ · 2026-09-26
- One amphibian question can tell if a model treats your prompt as a capability eval — AdtRaghunathan · 2026-09-26
- Perceptron launches Mk1.5, one embodied AI model for drones, quadrupeds and smart glasses — code_star · 2026-09-26
- Tester: Astra is "autistic" at parsing human sentiment; Fable and Opus run circles around it — teortaxesTex · 2026-09-26