EBench task scoreBrowse 284 families
EBench task score
Can the model plan, use tools, and finish a multi-step task?
Version / measure
- Domain
- Agents
- Ranked models
- 5
- Reported values
- 6
- Reports
- 1 · lab
- Direction
- Higher is better
- Status
- Version not specified
Rankings
Best reported score per model across effort settings
| # | Model | Best setting | Score |
|---|---|---|---|
| 1 | Qwen-RobotManipQwen | reportedcost n/a2 values · Jun 15 · source | 55 |
| 2 | InternVLA-A1External baseline | reportedcost n/a1 value · Jun 15 · source | 46 |
| 3 | π₀Physical Intelligence | reportedcost n/a1 value · Jun 15 · source | 41 |
| 4 | π₀.₅Physical Intelligence | reportedcost n/a1 value · Jun 15 · source | 39 |
| 5 | X-VLAExternal baseline | reportedcost n/a1 value · Jun 15 · source | 25 |
Effort curve
Score vs. reported cost. 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.
EBench's source-defined task score, kept separate from success rate.
Method / source ↗