Skip to content

Benchmark

OSWorld 2.0 · Partial

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

Models measured
24
Scores recorded
76
Sources
1
IQ weight
No weight published

Score distribution

59.3%
Mean
62.2%
Median
16.4%
Std dev
41.4%
10th percentile
75.4%
90th percentile
8.3%
Min
8.3%87.2%
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.87.2%
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.83.9%
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.81.8%
4Claude 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.80.7%
5Claude Sonnet 5.52026-09-10 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.80.1%
6Claude Fable 5.12026-08 task set · partial· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.77.9%
7Claude Fable 5.1System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted · partial· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.77.9%
8Claude Opus 52026-08 task set · partial· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.75.4%
9Claude Opus 5System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted · partial· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.75.4%
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.74%
11GPT-6 AstraOSWorld 2.0 offline set · v2026.08.08 · partial reward · OpenAI release chart · max reasoning · displayed cost $9.44/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.73.5%
12GPT-6 AstraOSWorld 2.0 offline v2026.08.08 · max effort · $9.0733 average cost/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.73.5%
13Claude Fable 52026-08 task set · partial· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.72.9%
14Claude Fable 5System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted · partial· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.72.9%
15GPT-6 Astrav2026.08.08 · offline set · partial score· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.72.6%
16Claude Haiku 5.582-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.72.4%
17GPT-6.1 SolOSWorld 2.0 offline set · v2026.08.08 · partial reward · OpenAI release chart · max reasoning · displayed cost $1.27/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.71.4%
18GPT-6 AstraOSWorld 2.0 offline set · v2026.08.08 · partial reward · OpenAI release chart · xhigh reasoning · displayed cost $7.49/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.71.3%
19GPT-6 AstraOSWorld 2.0 offline v2026.08.08 · xhigh effort · $7.1744 average cost/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.71.3%
20Claude Opus 5v2026.08.08 · offline set · partial score· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.70.2%
21Claude Opus 5OSWorld 2.0 offline v2026.08.08 · max effort · $24.11 average cost/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.70.2%
22Claude Opus 5OSWorld 2.0 offline v2026.08.08 · xhigh effort · $23.91 average cost/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.70.1%
23GPT-6 AstraOSWorld 2.0 offline v2026.08.08 · high effort · $6.6024 average cost/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.70%
24GPT-6 AstraOSWorld 2.0 offline set · v2026.08.08 · partial reward · OpenAI release chart · high reasoning · displayed cost $6.91/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.70%
25GPT-6.1 SolOSWorld 2.0 offline set · v2026.08.08 · partial reward · OpenAI release chart · high reasoning · displayed cost $0.96/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.69.6%
26GPT-6.1 SolOSWorld 2.0 offline set · v2026.08.08 · partial reward · OpenAI release chart · xhigh reasoning · displayed cost $1.05/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.69.4%
27GPT-6 AstraOSWorld 2.0 offline set · v2026.08.08 · partial reward · OpenAI release chart · medium reasoning · displayed cost $5.36/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.69.3%
28GPT-6 AstraOSWorld 2.0 offline v2026.08.08 · medium effort · $5.1009 average cost/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.69.3%
29Claude Opus 5max · OSWorld 2.0 v08.08 · 108 tasks · common internal framework · partial reward· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.68.3%
30Muse Spark 1.3max · OSWorld 2.0 v08.08 · 108 tasks · common internal framework · partial reward· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.66.9%
31GPT-6.1 SolOSWorld 2.0 offline set · v2026.08.08 · partial reward · OpenAI release chart · medium reasoning · displayed cost $0.77/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.66.8%
32GPT-5.6 SolOSWorld 2.0 offline v2026.08.08 · max effort · $7.7105 average cost/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.66.2%
33Claude Opus 5OSWorld 2.0 offline v2026.08.08 · high effort · $15.95 average cost/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.65.9%
34GPT-5.6 Solv2026.08.08 · offline set · partial score· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.65.7%
35GPT-6 SolOSWorld 2.0 offline v2026.08.08 · max effort · $3.2543 average cost/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.64.4%
36GPT-6 SolOSWorld 2.0 offline set · v2026.08.08 · partial reward · OpenAI release chart · max reasoning · displayed cost $3.37/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.64.4%
37GPT-5.6 Solmax · OSWorld 2.0 v08.08 · 108 tasks · common internal framework · partial reward· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.62.7%
38GPT-6 AstraOSWorld 2.0 offline set · v2026.08.08 · partial reward · OpenAI release chart · low reasoning · displayed cost $2.72/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.62.2%
39GPT-6 AstraOSWorld 2.0 offline v2026.08.08 · low effort · $2.5457 average cost/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.62.2%
40GPT-5.6 SolOSWorld 2.0 offline v2026.08.08 · xhigh effort · $5.9257 average cost/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.60.9%
41GPT-6 SolOSWorld 2.0 offline v2026.08.08 · xhigh effort · $2.2127 average cost/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.60.5%
42GPT-6 SolOSWorld 2.0 offline set · v2026.08.08 · partial reward · OpenAI release chart · xhigh reasoning · displayed cost $2.30/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.60.5%
43Claude Opus 5OSWorld 2.0 offline v2026.08.08 · medium effort · $12.67 average cost/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.60.3%
44GLM 5.3 FlashZ.ai Flash设置 / partial· Reported by Z.ai· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.59.1%
45Gemini 3.8 Flashpartial / batch tool enabled· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.59%
46GPT-6.1 SolOSWorld 2.0 offline set · v2026.08.08 · partial reward · OpenAI release chart · low reasoning · displayed cost $0.42/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.59%
47GPT-6 SolOSWorld 2.0 offline set · v2026.08.08 · partial reward · OpenAI release chart · high reasoning · displayed cost $1.71/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.58.3%
48GPT-6 SolOSWorld 2.0 offline v2026.08.08 · high effort · $1.6379 average cost/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.58.3%
49GPT-5.6 SolOSWorld 2.0 offline v2026.08.08 · high effort · $4.4613 average cost/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.56.5%
50Claude Opus 5OSWorld 2.0 offline v2026.08.08 · low effort · $9.88 average cost/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.55.2%

Showing 1–50 of 76 results.

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.