Benchmark
OSWorld
Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages.
- Models measured
- 29
- Scores recorded
- 31
- Sources
- 3
- IQ weight
- 1.34
Score distribution
- 54.5%
- Mean
- 62.3%
- Median
- 23.3%
- Std dev
- 23%
- 10th percentile
- 81.2%
- 90th percentile
- 5%
- Min
This benchmark has a weight of 1.34 for chat models, measured over 29 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Qwen3.8 Max· Steel.dev Agent Leaderboards data | 86.1% |
| 2 | Claude Fable 5· Steel.dev Agent Leaderboards data | 85% |
| 3 | Claude Opus 4.8· Steel.dev Agent Leaderboards data | 83.4% |
| 4 | MiMo V2.6 Pro· Steel.dev Agent Leaderboards data | 82% |
| 5 | Claude Sonnet 5· Steel.dev Agent Leaderboards data | 81.2% |
| 6 | MiMo V2.6 Flash· Steel.dev Agent Leaderboards data | 80.8% |
| 7 | Doubao Seed 2.1 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 78.8% |
| 8 | Claude Sonnet 4.6· Steel.dev Agent Leaderboards dataValues from 3 sources
| 78.5% |
| 9 | Doubao Seed 2.1 Turbo· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 76.4% |
| 10 | GPT 5.4· Steel.dev Agent Leaderboards data | 75% |
| 11 | Claude Opus 4.6· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 72.7% |
| 12 | Qwen3 VL 235B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.Values from 2 sources
| 66.7% |
| 13 | Claude Opus 4.5· Epoch AI Benchmarking Hub dataValues from 3 sources
| 66.3% |
| 14 | Kimi K2.5· Epoch AI Benchmarking Hub dataValues from 2 sources
| 63.3% |
| 15 | Claude Sonnet 4.5· Epoch AI Benchmarking Hub dataValues from 3 sources
| 62.9% |
| 16 | GLM 5V Turbo· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 62.3% |
| 17 | Doubao Seed 1.8· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 61.9% |
| 18 | Claude Haiku 4.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 50.7% |
| 19 | Claude Sonnet 4· Epoch AI Benchmarking Hub data | 43.9% |
| 20 | Qwen3 VL 32B Thinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 41% |
| 21 | Qwen3 VL 235Bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 38.1% |
| 22 | Claude Sonnet 3.7· Epoch AI Benchmarking Hub data | 35.8% |
| 23 | Qwen: Qwen3 VL 8B Thinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 33.9% |
| 24 | Qwen3 VL 8B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 33.9% |
| 25 | Qwen3 VL 32B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 32.6% |
| 26 | Computer Use Preview· Epoch AI Benchmarking Hub data | 31.3% |
| 27 | qwen3-vl-30bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 30.6% |
| 28 | qwen3-vl-30b· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. | 30.3% |
| 29 | OpenAI o3medium· Epoch AI Benchmarking Hub data | 23% |
| 30 | Qwen2.5 VL 72B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 3 sources
| 8.8% |
| 31 | Qwen 2.5 VL 32B Instruct· Steel.dev Agent Leaderboards data | 5.9% |
Data credits: Steel.dev Agent Leaderboards, MIT (licence) · Data by LLM Stats · Epoch AI Benchmarking Hub, CC BY 4.0 (licence)