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.
external benchmark
SkillsBench v1.1 official leaderboard (BenchFlow). Every model the catalog carries a published SkillsBench official leaderboard (BenchFlow) value for, ranked by that value.
BenchFlow's own SkillsBench table: each agent setup runs the 87 tasks of version 1.1 with Skills, the instruction folders an agent can load, and most also run them without. 21 models in the ModelScale catalog have a published value.
Each point is a model with a published SkillsBench official leaderboard (BenchFlow) value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
21 models with a published SkillsBench official leaderboard (BenchFlow) 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 | Agent tasks run up to three times each, with Skills and, for most setups, without |
|---|---|---|---|
| 1 | OpenAI | 67.3 | |
| 2 | Anthropic | 61.2 | |
| 2 | Anthropic |
What the benchmark measures and how its value is published, from the source’s own record.
SkillsBench official leaderboard (BenchFlow) (SkillsBench v1.1 official leaderboard (BenchFlow)) is an external benchmark published by SkillsBench leaderboard. BenchFlow's own SkillsBench table: each agent setup runs the 87 tasks of version 1.1 with Skills, the instruction folders an agent can load, and most also run them without. Its evaluation set is described by the source as 87 tasks in 8 domains (SkillsBench v1.1). The source rates its difficulty as multi-step agent tasks.
Each model's result is published by SkillsBench leaderboard as agent tasks run up to three times each, with Skills and, for most setups, without. ModelScale prints that value to 1 decimal place, exactly as published, and never converts or rescales it. This page ranks the 21 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 | 60.8 |
| 5 | Z.AI | 58.4 |
| 6 | Tencent | 55.9 |
| 7 | 54.6 |
| 8 | Anthropic | 54.1 |
| 9 | Moonshot AI | 54.0 |
| 10 | MiniMax | 53.0 |
| 11 | OpenAI | 51.7 |
| 12 | Anthropic | 50.2 |
| 13 | Anthropic | 49.0 |
| 14 | 48.2 |
| 15 | Anthropic | 47.2 |
| 16 | xAI | 41.7 |
| 17 | OpenAI | 41.4 |
| 18 | Anthropic | 36.2 |
| 19 | MiniMax | 34.9 |
| 20 | Anthropic | 30.1 |
| 21 | 20.1 |
Evidence key: ObservedLast good
Rows are ordered by the value SkillsBench leaderboard 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.