Arena launches agentic router Max: frontier performance at 37% lower cost
arena · x · 2026-10-14
Arena released Max, an agentic model router for Agent Arena, with key facts:
- Trained on 2.5M+ real-world Agent Arena sessions; routes requests by task complexity, cost, and type.
- Ranked #1 in live Agent Arena testing: +14.7% net improvement at $3.27 median cost per task, beating the top public model Claude Fable 5.1 (Max) by 0.3 points at 37.5% lower cost.
- The routing methodology generalizes across model combos: Fable 5 + Kimi K3 achieves +10.7% at $1.06 per task, outperforming Fable 5 (High) by 0.6 points while cutting cost 54%.
- Arena tested 60+ agentic models over the past 4 months to map performance-cost trade-offs across chat, code, and work.
- Max stays on the frontier as new models launch, suggesting the approach is model-agnostic.
More from Models
- Mystery model Fable 5.5 spotted in Azure catalog with GA status, launch imminent — koltregaskes · 2026-10-14
- Opus 5.5 builds Keynote decks via computer use, faster than Codex in this test — jdjohnson · 2026-10-14
- Claude picks 7 all 100 times: one prompt hack makes LLM output truly random — beglen · 2026-10-14
- Blind AI self-grading: Astra picked its own analysis #1, Muse ranked itself last — kevinkern · 2026-10-14
- Caught fabricating episode counts, Google's AI confesses: 'I was lying to you' — LoudStretch2997 · 2026-10-14
- Ramp spend data: Perplexity, Wispr Flow, Claude Code dominate AI buying categories — AravSrinivas · 2026-10-14