Agent Arena Leaderboard: Claude Opus 5 Tops Real-World Agentic Tasks, Kimi K3 at #4
arena · x · 2026-08-21
Agent Arena ranks models on millions of real-world, long-horizon agentic tasks from a global user community, with access to web search, filesystem and terminal tools, measured via causal tracing against the average model. Current leaders: Claude Opus 5 (High) tops with 12.47% net improvement (19,787 sessions, $2.13/task), followed by Opus 5 (Max) and Claude Fable 5; Kimi K3 (Max) ranks #4 (10.41%, 83,274 sessions, only $0.63/task — standout value), with GPT 5.6 Sol (xHigh) at #5. A user-signal update shows Muse Spark 1.2 (xHigh) gained +11.4% in Bash recovery and +6% confirmed success, but -2.5% steerability.
More from Models
- GLM 5.5 rumored to drop soon, coding ability expected to match frontier models — bindureddy · 2026-10-16
- Five rounds of adversarial prompting make Opus 5.5 and Argon converge on identical output — LanceCharger · 2026-10-16
- DIY benchmark ranks 18 models on subjective judgments, GLM-5.3 tops consensus match — arkuto · 2026-10-16
- Bindu Reddy: Google's Argon excels at complex doc writing at half the cost of Opus — bindureddy · 2026-10-16
- MAI-2 is 2x bigger than V4-Pro while V4.1F is 3x smaller: Microsoft's scaling ladder questioned — teortaxesTex · 2026-10-16
- Step 5 Preview hits DeepInfra day-zero: 600B MoE, 1M context, $1/M input tokens — StepFun_ai · 2026-10-16