Opus 5.5 (High) hits #2 in Agent Arena at 40-56% lower cost, new Pareto frontier
arena · x · 2026-09-29
LMArena reports Claude Opus 5.5 (High) debuts at #2 in Agent Arena (behind only Fable 5.1 Max) with a +12.15% net improvement score while costing 40-56% less than both Opus 5 variants:
- Opus 5.5 (High): $1.31 median per task / +12.15% net improvement
- Opus 5 (Max): $2.98 / +9.58%
- Opus 5 (High): $2.17 / +9.47%
By signal: #1 Steerability (+14.50%), #2 Confirmed Success (+15.50%), #3 Praise vs Complaint (+19.80%), #4 Bash Recovery (+10.64%). The release pushes the cost-performance Pareto frontier outward.
Related event: Claude Opus 5.5 Debuts Second on Agent Arena at 40-56% Lower Cost(2 posts)→
More from coding & agent
- DiffusionGemma becomes an open, self-hostable decision model on vLLM amid Jev's viral 'System One' moment — rseroter · 2026-09-29
- Hybrid Gemma + decision-head voice agent runs low-latency on Jetson tiers — cortexist · 2026-09-29
- Ex-Google DeepMind engineering lead launches Fo, a personal AI that employs humans to finish tasks — Scobleizer · 2026-09-29
- Scumble: free open-source inpainting editor with layers, full-res stitching and MCP support — Direction_Mountain · 2026-09-29
- Giving a coding agent persistent memory: retain/recall beats stuffing 8k tokens into the prompt — virat_gazab100 · 2026-09-29
- Refero turns 2,000+ real design systems into DESIGN.md files so coding agents stop making ugly UI — victor_explore · 2026-09-29