SkillsBench without Skills
Agents · 2026-02-13
SkillsBench resolution rate for the no-Skills condition on the 87-task v1.1 suite. The current site leaderboard publishes this paired ablation for 24 model-harness configurations, with up to three trials per task.
Top models (higher is better)
| Model | Score |
|---|---|
| DeepSeek V4.1 Flash | 61.7 |
| GPT-5.5 | 51.5 |
| GPT-5.4 | 48.9 |
| GPT-5.6 Terra | 48.8 |
| Grok 4.5 | 48.8 |
| Muse Spark 1.1 | 46.3 |
| Opus 4.8 | 45.7 |
| GPT-5.6 Luna | 45.2 |
| Gemini 3.7 Flash | 44.9 |
| GPT-5.6 Sol | 43.3 |
| Opus 4.7 | 43.0 |
| DeepSeek V4 Pro 0813 | 42.1 |
| Gemini 3.5 Flash | 41.1 |
| Gemini 3.8 Flash | 41.0 |
| Muse Spark 1.2 | 40.6 |
| DeepSeek-V4-Flash-0731 | 38.3 |
| Grok 4.6 | 38.2 |
| Gemini 3.1 Pro Preview | 36.0 |
| Sonnet 5 | 35.1 |
| Gemini 3 Flash Preview | 34.2 |
| Opus 4.6 | 33.7 |
| Sonnet 4.6 | 33.5 |
| Kimi K2.6 | 33.4 |
| GLM-5.1 | 32.7 |
| GLM-5.3 Flash | 32.5 |
| GLM-5.2 | 32.2 |
| Qwen3.7-Plus (source unspecified) | 31.1 |
| Kimi K2.7 Code | 30.4 |
| Qwen3.8-Max | 30.3 |
| GPT-5.4 Mini | 29.9 |
| MiniMax M3 | 29.7 |
| GPT-5.2-Codex | 29.7 |
| GLM-5.3 | 29.6 |
| DeepSeek-V4-Flash | 27.5 |
| DeepSeek-V4-Pro | 27.2 |
| Opus 4.5 | 23.8 |
| Grok 4.3 | 22.8 |
| Inkling-Small | 21.9 |
| Qwen3.8 27B | 19.1 |
| Inkling | 18.5 |