DeepSearchQABrowse 284 families
DeepSearchQA
Can the model plan, use tools, and finish a multi-step task?
Version / measure
- Domain
- Agents
- Ranked models
- 7
- Reported values
- 8
- Reports
- 2 · lab
- Direction
- Higher is better
- Status
- Version not specified
Rankings
Best reported score per model across effort settings
| # | Model | Best setting | Score |
|---|---|---|---|
| 1 | Kimi K3Moonshot AIOW | maxcost n/a1 value · Jul 17 · source | 95.0% |
| 2 | Claude Fable 5Anthropic | maxcost n/a1 value · Jul 17 · source | 94.2% |
| 3 | Claude Opus 4.8Anthropic | maxcost n/a2 values · Jul 17 · source | 93.1% |
| 4 | GPT-5.5OpenAI | xhighcost n/a1 value · Jul 9 · source | 87.8% |
| 5 | Muse Spark 1.1Meta | xhighcost n/a1 value · Jul 9 · source | 84.9% |
| 6 | Muse Spark 1.0Meta | reportedcost n/a1 value · Jul 9 · source | 76.8% |
| 7 | Gemini 3.1 ProGoogle | highcost n/a1 value · Jul 9 · source | 71.3% |
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. The value is the percentage reported in this lab's table.
Harness, tools, prompts, attempt count, and benchmark version can materially change the result. Compare within one reporting context.
Browser search/open/find harness with gpt-oss-120B judge.
Method / source ↗