Artificial Analysis Coding Agent Index v2.1: Claude Models Tie for Lead at 58
ArtificialAnlys · x · 2026-11-14
Artificial Analysis shipped Coding Agent Index v2.1, upgrading FrontierCode to AA 2.0 (132 real-world tasks graded by whether maintainers would merge) and Terminal-Bench to 4.1 with anti-reward-hacking verifiers.
- Claude Fable 5.2 and Claude Opus 5.5 tie for the lead at 58; Claude Sonnet 5.5 (56) and Gemini 4 Carbon (55) follow
- Scores dropped 5 points on average under the harder v2.1 benchmarks; Codex with GPT-6.1 Terra was the only config within 1 point of its prior score
- Cost per task among 48+ scoring configs varies 30x: $1.51 (Codex + GPT-6.1 Sol, 52) to $44.22 (Claude Code + Fable 5.2 max, 56)
Related event: Artificial Analysis Coding Agent Index v2.1: Claude Duo Tops Chart(3 posts)→
More from Models
- Dev builds tabletop puzzle game combining this week's Gemini model launches — gaganghotra_ · 2026-11-14
- Anthropic Breaks Down the Five Primitives Behind Always-On AI Agents — dotey · 2026-11-14
- Model trolls its developer with fake mid-stream truncations during harness fix — repligate · 2026-11-14
- Gemini 4 Hydrogen writes 3.7x faster yet takes 474s per task, matching GPT-6.1 Sol — allenainie · 2026-11-14
- Reward hacking rates compared: Mirror-Beam hits 0.51% while most frontier coding models score zero — teortaxesTex · 2026-11-14
- Users worldwide report ChatGPT suddenly shorter, colder, memory broken — ShadowNelumbo · 2026-11-14