Benchmark
OSWorld 2.0
Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages. Higher is better.
- Models measured
- 21
- Scores recorded
- 28
- Sources
- 2
- IQ weight
- 1.81
Score distribution
- 58.2%
- Mean
- 61.7%
- Median
- 14.3%
- Std dev
- 45.6%
- 10th percentile
- 71.4%
- 90th percentile
- 19.4%
- Min
This benchmark has a weight of 1.81 for chat models, measured over 18 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 | Claude Opus 5.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 81.8% |
| 2 | Claude Fable 5.1· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 77.9% |
| 3 | GPT-6 Astra· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 72.6% |
| 4 | GPT-6.1 Sol· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 71.4% |
| 5 | Claude Opus 5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 70.6% |
| 6 | Claude Opus 5Opus 5 System Card setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 70.6% |
| 7 | Gemini 4 Argon· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 69.2% |
| 8 | Muse Spark 1.3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 66.9% |
| 9 | Claude Fable 5Opus 5 System Card setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 66.1% |
| 10 | GPT-6 Sol· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 64.4% |
| 11 | GPT-5.6 Sol· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 62.6% |
| 12 | GPT-5.6 SolOpus 5 System Card setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 62.6% |
| 13 | GPT-5.6 SolOSWorld 2.0 · OpenAI GPT-5.6 launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 62.6% |
| 14 | MiniMax M3.1 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 61.7% |
| 15 | MiniMax M3.1 FlashMiniMax M3.1 Flash official-weights model-card run · reasoning effort, sampling, and run count not stated for this benchmark · OSWorld 2.0 0808 Full release · highest reported at release · metric subtype unstated· Reported by MiniMax· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 61.7% |
| 16 | Gemini 3.8 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 59% |
| 17 | Kimi K3max effort · temp 1.0 · top_p 1.0· Reported by Moonshot AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 58.3% |
| 18 | Claude Opus 4.8Opus 5 System Card setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 55.7% |
| 19 | Claude Opus 4.8OSWorld 2.0 · OpenAI GPT-5.6 launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 54.8% |
| 20 | GPT-6 Luna· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 52.7% |
| 21 | GPT-5.6 Terra· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 50.2% |
| 22 | GPT-5.6 TerraOSWorld 2.0 · OpenAI GPT-5.6 launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 50.2% |
| 23 | Gemini 3.7 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 47.9% |
| 24 | GPT 5.5OSWorld 2.0 · OpenAI GPT-5.6 launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 47.5% |
| 25 | GPT-5.6 Luna· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 45.6% |
| 26 | GPT-5.6 LunaOSWorld 2.0 · OpenAI GPT-5.6 launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 45.6% |
| 27 | Qwen 3.8 Flash Next· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 19.4% |
| 28 | Qwen3.8 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 19.4% |
Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · Data by LLM Stats