Agent Arena: open models match Claude Fable 5 in 12 weeks at up to 96% lower cost
arena · x · 2026-10-13
Arena launched Agent Arena, ranking agents by "Confirmed Success" — users verifying real delegated tasks. Claude Fable 5 led on July 1, but the top-20 average matched it within 12 weeks. Kimi K3 beat it in 3 weeks at 89% lower cost per task; DeepSeek V4.1 Flash matched at 96% lower (GLM 5.3 Flash at 95%). Top-5 confirmed success rose to 94% from 80% in July, led by Claude Opus 5.5, Claude Fable 5.1 and GPT-6.1 Sol. Agent leadership is commoditizing fast.
Related event: Agent Arena Launches Agent Reliability Leaderboard(2 posts)→
More from Models
- Google AI Overviews denied a person had died, renewing criticism of AIO for news — lilyraynyc · 2026-10-13
- User feeds Claude Opus GTA frame-by-frame screenshots since it can't watch videos — adonis_singh · 2026-10-13
- Gemini Deep Research on web is really slow today, user reports — gaganghotra_ · 2026-10-13
- Dual RAID10 NVFP4 setup loads GLM 5.3 Flash cold in ~1 minute — TheZachMueller · 2026-10-13
- Opus 5.5 stuns at SVG generation; Meta's Muse Wonder judged 3-6 months behind frontier — scaling01 · 2026-10-13
- Muse vs Opus 5.5: same 3D prompt costs $4.11 vs $99.43, big quality gap — iamaliveix · 2026-10-13