GPT-6 Astra reshapes Agent Arena Pareto frontier at $4.01/task for +12.55%
arena · x · 2026-09-09
Arena's Agent Arena leaderboard measures net improvement on real-world, long-horizon tasks using web, filesystem, and terminal tools. GPT-6 Astra (Max) has reshaped the Pareto frontier at $4.01/task for +12.55% net improvement; paying 7% more ($4.30) gets Claude Fable 5.1 (Max) at +14.5%, two points higher. Below $4.01, no model scores higher.
Unlike voting-based leaderboards, Agent Mode routes real sessions to random models and scores them as orchestrators across millions of in-the-wild interactions.
Related event: GPT-6 Astra Reshapes Agent Arena Pareto Frontier at $4.01/Task(2 posts)→
More from Models
- Benchmark scores drop from 89% to 19% on new evals — how benchmaxxing breaks leaderboard trust — airesearch12 · 2026-09-11
- Developer Building a Unified Leaderboard of All Model Benchmark Scores — airesearch12 · 2026-09-11
- Rumor claims Kimi faked performance by serving Claude; DeepSeek new model surprises in evals — realsohamparekh · 2026-09-11
- GPT-5.6 writes well but is instantly forgettable, user complains — BasedRaddka · 2026-09-11
- Opus Refuses Protein Research Codebase Over 'Safety' Concerns, Dev Considers Rolling His Own — josephdviviano · 2026-09-11
- User Hails Unconfirmed 'DeepSeek 4.1 Flash' as an Inflection Point in LLMs — himanshustwts · 2026-09-11