Skip to content

Benchmark

OSWorld 2.0 · Strict

Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages. Higher is better.

Models measured
6
Scores recorded
13
Sources
1
IQ weight
No weight published

Score distribution

42%
Mean
41.7%
Median
5.2%
Std dev
36.1%
10th percentile
48.8%
90th percentile
36.1%
Min
36.1%53.2%
The band covers the middle 80% of scores. The line marks the median.

Leaderboard

#ModelScore
1Claude Opus 5.5adaptive thinking · max effort · default sampling · 5-trial mean · context ≤1M unless noted · 82-task offline subset · max · 5 runs· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.53.2%
2Claude Sonnet 5.5adaptive thinking · max effort · default sampling · 5-trial mean · context ≤1M unless noted · 82-task offline subset · max · 5 runs· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.48.8%
3Claude Opus 5.5Opus 5.5 System Card capability section · 2026-09-10 task/assets · 1080p · 500 steps · max · 5 runs· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.48.7%
4Claude Sonnet 5.5adaptive thinking · max effort · default sampling · 5-trial mean · context ≤1M unless noted · 2026-09-10 tasks/assets · 1080p · 500 steps · max · 5 runs· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.43.5%
5Claude Fable 5.1Opus 5.5 System Card capability section · 2026-09-10 task/assets · updated harness · max · 5 runs· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.42.8%
6Claude Fable 5.12026-08 task set · strict· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.41.7%
7Claude Fable 5.1System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted · strict· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.41.7%
8Claude Opus 52026-08 task set · strict· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.39.6%
9Claude Opus 5System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted · strict· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.39.6%
10Claude Opus 5Opus 5.5 System Card capability section · 2026-09-10 task/assets · updated harness · max · 5 runs· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.37.2%
11Claude Haiku 5.5adaptive thinking · max effort · default sampling · 5-trial mean · context ≤1M unless noted · 82-task offline subset · 1080p · 500 steps · max · 5 runs· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.37.1%
12Claude Fable 52026-08 task set · strict· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.36.1%
13Claude Fable 5System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted · strict· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.36.1%

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.