PBench
Can the model plan, use tools, and finish a multi-step task?
Agents11 ranked models11 reported values1 reportsHigher is better
Top rankings
One row per model, using its best reported score across effort settings.
| Rank | Model | Best score | Best reported setting |
|---|---|---|---|
| 1 | Veo 3Google | 0.526 | Reported configurationQwen report1 value · 1 reportJun 15, 2026 · source |
| 2 | KlingExternal baseline | 0.521 | Reported configurationQwen report1 value · 1 reportJun 15, 2026 · source |
| 3 | WowExternal baseline | 0.517 | Reported configurationQwen report1 value · 1 reportJun 15, 2026 · source |
| 4 | LVPExternal baseline | 0.515 | Reported configurationQwen report1 value · 1 reportJun 15, 2026 · source |
| 5 | Wan2.6External baseline | 0.514 | Reported configurationQwen report1 value · 1 reportJun 15, 2026 · source |
| 6 | LTX-2External baseline | 0.506 | Reported configurationQwen report1 value · 1 reportJun 15, 2026 · source |
| 7 | VidarExternal baseline | 0.501 | Reported configurationQwen report1 value · 1 reportJun 15, 2026 · source |
| 8 | GigaWorldExternal baseline | 0.495 | Reported configurationQwen report1 value · 1 reportJun 15, 2026 · source |
| 9 | Sora2OpenAI | 0.487 | Reported configurationQwen report1 value · 1 reportJun 15, 2026 · source |
| 10 | CosmosNVIDIA | 0.470 | Reported configurationQwen report1 value · 1 reportJun 15, 2026 · source |
| 11 | Qwen-RobotWorldQwen | 0.455 | Reported configurationQwen report1 value · 1 reportJun 15, 2026 · source |
Effort curve
Every sourced cost-linked effort value for this exact version. Lines connect complete sweeps only.
No cost-linked effort sweep for this version.
Definition and comparison boundarypublic methodology
Can the model plan, use tools, and finish a multi-step task?
Agentic execution across tools, environments, or long-running workflows. Higher is better. This benchmark reports points rather than percent correct.
Harness, tools, prompts, attempt count, and benchmark version can materially change the result. Compare within one reporting context.
PBench source-defined quality, consistency, motion, and domain metrics.
Method / source