Agent Arena: Top Agent Models Differ Over 5x in Cost per Task
arena · x · 2026-08-20
Agent Arena published an agent-model value ranking based on "net improvement × cost per task" from real agentic sessions (1.87M sessions, 49 models).
- Pareto-optimal: Claude Opus 5 (High) leads at $1.78/task, +12.34%; Kimi K3 (Max) is the value dark horse at $0.62/task, +10.53%; GPT 5.5 (High) at $0.44/task, +7.75%; Grok 4.5 at $0.22/task, +6.08%; GLM 5.2 (Max) at $0.18/task, +6.06%.
- Even within the top 5, median cost per task ranges from $0.62 (Kimi K3 Max) to $3.37 (Claude Opus 5 Max) — a >5x difference.
- Claude Opus 5 (Max) costs nearly twice as much as Opus 5 (High) while scoring slightly lower (+11.97% vs +12.34%).
More from Models
- Meta open-sources Muse Spark Open agent model with day-0 TPU support in SGLang — BanghuaZ · 2026-10-16
- Saying "load-bearing" in your prompt makes Opus 5.5 stop holding back — zacharynado · 2026-10-16
- Americans get their Claude accounts banned while on business trips in China — dotey · 2026-10-16
- I priced every Grok Bot call: $20 to $200 plan in 5 days, long jobs eat 52% — jonathanmalkin · 2026-10-16
- NVIDIA releases Nemotron-3.5-Embed-1B, a multimodal embedding model covering 34 languages — jacek2023 · 2026-10-16
- EpochAI pegs GPT-6 training at up to 1.8e27 FLOP; scaling01 calls param estimates way too high — scaling01 · 2026-10-16