Agent Arena leaderboard: Claude Opus 5.5 tops net improvement across 2.38M sessions
arena · x · 2026-10-13
Arena shared the full Agent Arena leaderboard (2.38M+ sessions, 53 models): Claude Opus 5.5 (Max) leads with 15.53% net improvement, followed by GPT 6 Astra (13.09%) and Claude Fable 5.1 (12.66%). GPT 6.1 Sol is cheapest at $0.62/task; Gemini 4 Argon has the lowest tool hallucination (0.18%). The board tracks steerability, bash recovery and other agent signals in real time.
Related event: Agent Arena Launches Agent Reliability Leaderboard(2 posts)→
More from Models
- Meta's Alexandr Wang says natural communication was a big focus for Muse Wonder — alexandr_wang · 2026-10-13
- Dev says Opus 5.5 now feels sloppy and needs hand-holding like every other LLM — audrow · 2026-10-13
- Decision models: the new Hub category that outputs numbers, not paragraphs — erhannah · 2026-10-13
- Muse Wonder Video Model Tested: 'Pretty Rough' Output, Every Narration in Australian Accent — danshipper · 2026-10-13
- Top 10 AI models' share of developer tokens plunges from 55% to 14% — GaryMarcus · 2026-10-13
- DeepSeek V4.1 Flash suspected silent update shows high code-edit failure rate — teortaxesTex · 2026-10-13