Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Agentic benchmark
OSWorld 2.1 (all 108 tasks). Every model the catalog carries a published OSWorld 2.1 value for, ranked by that value.
OSWorld 2.1 is the current release of OSWorld 2.0: the same 108 long-horizon computer-use workflows with the maintainers' later fixes to task files, assets, mocked websites, and evaluators. This page stores partial scores on the full task set. 12 models in the ModelScale catalog have a published value.
Each point is a model with a published OSWorld 2.1 value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
12 models with a published OSWorld 2.1 value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Interactive computer-use evaluation; partial score |
|---|---|---|---|
| 1 | Anthropic | 81.8 | |
| 2 | Anthropic | 80.7 | |
| 3 | Anthropic | 80.1 |
What the benchmark measures and how its value is published, from the source’s own record.
OSWorld 2.1 (OSWorld 2.1 (all 108 tasks)) is an agentic benchmark published by OSWorld 2.1 release (xlang-ai/OSWorld-V2, tag osworld-v2.1). OSWorld 2.1 is the current release of OSWorld 2.0: the same 108 long-horizon computer-use workflows with the maintainers' later fixes to task files, assets, mocked websites, and evaluators. This page stores partial scores on the full task set. Its evaluation set is described by the source as 108 long-horizon computer-use workflows. The source rates its difficulty as long-horizon professional workflows.
Each model's result is published by OSWorld 2.1 release (xlang-ai/OSWorld-V2, tag osworld-v2.1) as interactive computer-use evaluation; partial score. ModelScale prints that value to 1 decimal place, exactly as published, and never converts or rescales it. This page ranks the 12 models that carry a published value, highest value first; a model the source never scored is left out rather than ranked at zero. The source does not state whether a higher value is the better result, so neither does this page.
| 4 | Moonshot AI | 79.6 |
| 5 | Tencent | 78.9 |
| 6 | Meta | 78.4 |
| 7 | OpenAI | 75.8 |
| 8 | Anthropic | 74.0 |
| 9 | Meta | 72.6 |
| 10 | Meta | 69.5 |
| 11 | Tencent | 68.1 |
| 12 | Anthropic | 57.0 |
Evidence key: Observed
Rows are ordered by the value OSWorld 2.1 release (xlang-ai/OSWorld-V2, tag osworld-v2.1) published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.