Which model leads AgentWorldBench - Web?

Q: Which model leads AgentWorldBench - Web?

Claude Opus 4.8 tops the board at 54.66

AgentWorldBench - Web Leaderboard — 14 Models

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

The field · 14 measured

47.154.66

14 models measured, most on the third-party eval harness. Claude Opus 4.8 tops the board at 54.66.

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 47.1–54.66 · ◆ solid = first-party or better · outlined = arm's-length

01Claude Opus 4.8Anthropic54.66$5.0/M$25.0/M$0.274

02Gemini 3.1 ProGoogle52.83$2.0/M$12.0/M$0.133

03GPT-5.4OpenAI51.8$2.5/M$15.0/M$0.169

04GLM-5.1Z.ai51.5$1.4/M$4.4/M$0.056

05Claude Opus 4.6Anthropic51.42$5.0/M$25.0/M$0.292

06Qwen AgentWorld 397B A17BAlibaba50.98———

07Qwen3.6 PlusAlibaba50.78$0.50/M$3.0/M$0.034

08Claude Sonnet 4.6Anthropic50.78$3.0/M$15.0/M$0.177

09MiniMax M2.7MiniMax50.52$0.21/M$0.84/M$0.010

10DeepSeek-V4-ProDeepSeek50.32$0.43/M$0.87/M$0.013

11Kimi K2.6Moonshot50.2$0.95/M$4.0/M$0.049

12Qwen AgentWorld 35B A3BAlibaba49.55———

13Qwen3.5 397B A17BAlibaba48.55$0.60/M$3.6/M$0.043

14Qwen3.5 35B A3BAlibaba47.1$0.25/M$2.0/M$0.024

◆ 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