Artificial Analysis launches Coding Agent Index comparing 39 model-agent combos
ArtificialAnlys · x · 2026-10-14
Artificial Analysis published the methodology and full leaderboard for its Coding Agent Index v1.5, an equal-weight composite of DeepSWE v1.1 (113 tasks, Datacurve), Terminal-Bench 4.0 (66 tasks, Laude Institute) and SWE-Atlas-QnA (124 tasks, Scale AI), averaging pass@1 over three attempts per task. The site provides per-task wall time, pay-per-token API cost, and per-benchmark breakdowns across 39 model/agent combinations, with a 'most attractive quadrant' and Pareto frontier view for cost-performance comparisons across agents, models and effort settings.
More from coding & agent
- With superhuman coding agents, changing habits is still the hardest engineering problem — daniel_mac8 · 2026-10-14
- Indie tool turns design files into code, works with any agent unlike Figma MCP — talkaboutdesign · 2026-10-14
- Robo Use launches: one API for agents across 32 robot embodiments — joemeno · 2026-10-14
- 10 Open-Source GitHub Repos That Give Your AI Agent Memory, from Mem0 (67K Stars) to Engram — victor_explore · 2026-10-14
- Free Bots: A Persistent 3D City Where AI Agents Actually Live, Work, and Build Houses — Daniel_Farinax · 2026-10-14
- Caught in the logs: Codex quietly delegated work to DeepSeek Flash — dotey · 2026-10-14