Arena's agentic Max router, trained on 2.5M Agent Arena sessions, tops Agent Arena at a $3.27 median task cost — 37.5% below Fable 5.1 (Max)
Arena released an agentic version of its Max router. Trained on 2.5M+ real Agent Arena sessions, it picks between Fable 5.1 (Max), Astra (Max) and Astra (Medium) per task based on complexity, task type and cost. In a live test it ranked #1 in Agent Arena with a +14.7% score at a $3.27 median cost per task — 0.3% better than the top public model, Fable 5.1 (Max), at 37.5% lower cost — and sent 67% of coding sessions to the cheaper Astra (Medium). It is free to try in Agent Mode for a limited time; API access is via a signup form.
Key Takeaways
- Training data: 2.5M+ real Agent Arena sessions; Arena tested 60+ agentic models over the past 4 months
- Live test: #1 in Agent Arena with +14.7% score and a $3.27 median cost per task — 0.3% above Fable 5.1 (Max) at 37.5% lower cost
- Routing mix: code 67% → Astra (Medium); chat Astra (Max) 49% / Fable 5.1 (Max) 42%; work Fable 5.1 (Max) 49% / Astra (Medium) 43%
- Model-agnostic: an earlier Fable 5 + Kimi K3 router hit +10.7% net at a $1.06 median cost, 0.6 pp above Fable 5 (High) at 54% lower cost
- Caveats: trained on data collected before Sept 17 (pre-Opus 5.5); Agent Mode only for now, API via signup
Linked Models & Agents:
