Claude Fable 5.1 tops Agent Arena at $4.14/task median cost with +15.8% net gain
arena · x · 2026-09-06
- Agent Arena (LMArena) update: Anthropic's Claude Fable 5.1 (Max) takes the #1 spot with a +15.8% net improvement, redrawing the price-performance Pareto frontier — it is both the best-performing and the most expensive model, at a median cost of $4.14 per task.
- Scale: based on 6.7k+ real-world agentic sessions.
- By signal: implicit user sentiment (Praise vs. Complaint) at +42.5% — users praise it roughly 2x more than the next top model; Confirmed Success +22.4%; Bash Recovery +13.1%; zero Tool Hallucinations.
- The arena says traces for GPT-6 Astra are coming to enable comparison; users can contribute via Agent Mode.
Related event: Claude Fable 5.1 Tops Agent Arena Leaderboard(3 posts)→
More from coding & agent
- Dev demos AI models handling 3D objects and animations surprisingly well — chongdashu · 2026-10-23
- Jason Kneen burns 104.56B tokens and $24.7K in 304 days of AI coding on Codex and Claude Code — jasonkneen · 2026-10-23
- Typosquatted Blender MCP Repo Hit 958 Stars in Days and Silently Ran Malware — Promptmethus · 2026-10-23
- Benchmarking 56 fresh dependency bugs: MCP routing lifts Claude Code accuracy 68% to 91% — Available_Side_7134 · 2026-10-23
- Swapping only the eval harness moves results 30%: designing agent evals like a scientist — AI Engineer · 2026-10-23
- Med student builds a whole-library RAG with Cognee; single-book ingestion still freezes his machine — shrike293103 · 2026-10-23