Artificial Analysis Releases Capability Indices v1.1… Claude Fable 5.1 Tops All 6 Categories, Grok 4.5 Noted for Cost Efficiency

Artificial Analysis released Capability Indices v1.1 and Intelligence Index v4.2, re-evaluating the performance and cost of the latest AI models. Claude Fable 5.1 ranked first in all six Capability Indices: Finance & Accounting, Strategy & Ops, Legal, Healthcare & Medical, Engineering, and Economics. Among open-weight models, Kimi K3 took first place in Finance & Accounting, Legal, and Economics; DeepSeek V4.1 Flash in Strategy & Ops; and GLM-5.3 in Healthcare & Medical and Engineering.

Grok 4.5 demonstrated strong performance in agentic tasks, ranking 4th (Elo 1543) on GDPval-AA v2 and scoring 33% on 𝜏³-Banking, surpassing GPT-5.5 (31%). In terms of cost, it recorded a headline price of $0.31 per task based on the Intelligence Index and $2.49 based on the Coding Agent Index (Grok Build), making it over 60% cheaper than Opus 4.8 ($11.80) and positioning it on the cost-performance Pareto frontier.

Gemini 3.1 Pro Preview ranked first in 6 of the 10 evaluations comprising the Intelligence Index (including Terminal-Bench Hard, AA-Omniscience, etc.) and can run at less than half the cost of leading models such as Opus 4.6 and GPT-5.5. However, it costs approximately twice as much as leading open-weight models like GLM-5. Additionally, the newly released Endpoint Accuracy Index measures the accuracy preservation rates of third-party hosting for GLM-5.2, gpt-oss-120b, and DeepSeek V4 Pro.