Which model leads AgentWorldBench - OS?

Q: Which model leads AgentWorldBench - OS?

Claude Opus 4.6 tops the board at 70.2

AgentWorldBench - OS Leaderboard — 14 Models

Pro Mode — raw scores, one benchmark under the lensthird-party eval harness

The field · 14 measured

56.2770.2

14 models measured, most on the third-party eval harness. Claude Opus 4.6 tops the board at 70.2.

14 measured·Lifecycle Updated Oct 8, 2026Score basis

#◆Vendor$/M in$/M out

Leader — ignites whiteField — fill cools with scoreIndependently verified only0 of 14Bars show 56.27–70.2 · ◆ solid = first-party or better · outlined = arm's-length

01Claude Opus 4.6Anthropic70.2$5.0/M$25.0/M$0.214

02GPT-5.4OpenAI68.58$2.5/M$15.0/M$0.128

03Qwen AgentWorld 397B A17BAlibaba67.89———

04Gemini 3.1 ProGoogle66.92$2.0/M$12.0/M$0.105

05Claude Opus 4.8Anthropic66.62$5.0/M$25.0/M$0.225

06Qwen AgentWorld 35B A3BAlibaba65.92———

07DeepSeek-V4-ProDeepSeek63.7$0.43/M$0.87/M$0.010

08Claude Sonnet 4.6Anthropic63.17$3.0/M$15.0/M$0.142

09Qwen3.5 397B A17BAlibaba60.85$0.60/M$3.6/M$0.035

10Kimi K2.6Moonshot60.8$0.95/M$4.0/M$0.041

11Qwen3.6 PlusAlibaba60.33$0.50/M$3.0/M$0.029

12GLM-5.1Z.ai59.13$1.4/M$4.4/M$0.049

13MiniMax M2.7MiniMax57.73$0.21/M$0.84/M$0.009

14Qwen3.5 35B A3BAlibaba56.27$0.25/M$2.0/M$0.020

◆ solid = independently verified · aggregator attested · vendor attributed — outlined = cross-reference · source-attributed. Dimmed ◆ = confidence inferred. “+N alts” = alternate disclosures in the drawer with why-demoted. Prices per million tokens; $/point = input price ÷ score. Every row opens its model dossier.

Verification: 14 models scored · 0 independently verified · 9 vendor cross-reference · 5 vendor-reported · 0 with source disagreement. How these tiers are assigned