Skip to content

Benchmark

OSWorld-Verified

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

Models and systems measured
40
Scores recorded
81
Sources
2
IQ weight
2.20

Score distribution

73.2%
Mean
76.2%
Median
10.6%
Std dev
58%
10th percentile
83.4%
90th percentile
39%
Min
39%86.1%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.20 for chat models, measured over 38 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#Model or systemScore
1Qwen3.8 Max· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim86.1%
2Qwen3.8 MaxQwen3.8-Max release vision/video table · published setting· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.86.1%
3Claude Fable 5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim85%
4Claude Fable 5Fable/Mythos 5 System Card setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.85%
5Kimi 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.84.8%
6Qwen 3.8 27B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim84.3%
7Qwen 3.8 27BQwen3.8-27B model card · published setting· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.84.3%
8Claude Opus 4.8· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim83.4%
9Claude Opus 4.8Fable/Mythos 5 System Card setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.83.4%
10Claude Opus 4.8max · 361-task split excluding Google Drive · GUI-only · 1920×1080 · 200 steps · pass@1· Reported by Meta, Meta Superintelligence Labs· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.83.4%
11Claude Opus 4.8Opus 4.8 System Card setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.83.4%
12Gemini 3.6 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim83%
13Gemini 3.6 FlashGemini 3.6 Flash Model Card · July 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.83%
14Claude Opus 4.7Opus 4.8 System Card setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.82.8%
15MiMo V2.6 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim82%
16Claude Sonnet 5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim81.2%
17Claude Sonnet 5361 tasks · 1080p · 100 steps · pass@1 · 5 runs· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.81.2%
18Claude Sonnet 5Gemini 3.6 Flash Model Card · July 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.81.2%
19MiMo V2.6 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim80.8%
20Muse Spark 1.1· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim80.8%
21Muse Spark 1.1Meta Model API · xhigh · 361-task split excluding Google Drive · GUI-only · 1920×1080 · 200 steps · pass@1· Reported by Meta, Meta Superintelligence Labs· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.80.8%
22Muse Spark Open· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim80.4%
23GPT 5.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim78.7%
24GPT 5.5Fable/Mythos 5 System Card setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.78.7%
25GPT 5.5Gemini 3.5 Flash Model Card · May 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.78.7%
26GPT 5.5Opus 4.8 System Card setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.78.7%
27GPT 5.5xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.78.7%
28GPT 5.5xhigh · 361-task split excluding Google Drive · GUI-only · 1920×1080 · 200 steps · pass@1· Reported by Meta, Meta Superintelligence Labs· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.78.7%
29Claude Sonnet 4.6zoom-tool fix + 128K per-turn output · 361 tasks · 5 runs· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.78.5%
30Gemini 3.5 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim78.4%
31Gemini 3.5 FlashFable/Mythos 5 System Card setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.78.4%
32Gemini 3.5 FlashGemini 3.5 Flash Model Card · May 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.78.4%
33Gemini 3.5 FlashGemini 3.6 Flash Model Card · July 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.78.4%
34Gemini 3.5 FlashOpus 4.8 System Card setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.78.4%
35Claude Opus 4.7· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim78%
36Claude Opus 4.71080p · max 100 actions · first-attempt success · updated Anthropic scaffold · avg@5· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.78%
37Claude Opus 4.7Gemini 3.5 Flash Model Card · May 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.78%
38Gemini 3.1 Pro PreviewFable/Mythos 5 System Card setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.76.2%
39Gemini 3.1 Pro PreviewGemini 3.5 Flash Model Card · May 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.76.2%
40Gemini 3.1 Pro PreviewGemini 3.6 Flash Model Card · July 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.76.2%
41Gemini 3.1 Pro Previewhigh · 361-task split excluding Google Drive · GUI-only · 1920×1080 · 200 steps · pass@1· Reported by Meta, Meta Superintelligence Labs· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.76.2%
42Gemini 3.1 Pro PreviewOpus 4.8 System Card setting· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.76.2%
43MiniMax M3official model-page overview figure · setting not stated· Reported by MiniMax· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.75.2%
44GPT 5.4· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim75%
45GPT 5.4OpenAI GPT-5.4 launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.75%
46Gemini 3.5 Flash Lite· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim74%
47GPT-5.3 CodexOpenAI GPT-5.4 launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.74%
48Qwen3.7 Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim73.3%
49Qwen3.7 PlusQwen3.7-Plus release · enable_thinking=false· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.73.3%
50Kimi K2.6· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim73.1%

Showing 1–50 of 79 results.

Not ranked

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
Muse Spark (version unspecified)Meta Model API · highest effort · version unspecified · 361-task split excluding Google Drive · GUI-only · 1920×1080 · 200 steps · pass@1· Reported by Meta, Meta Superintelligence Labs· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.53.3%
Qwen3.8-2.4T-A95Blifecycle-page core-benchmark claim· Reported by Alibaba Cloud Model Studio· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.86.1%

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. · Data by LLM Stats