# Vals SkillsBench (SkillsBench)

> How important are skills for agents?

Canonical page: https://benchlm.ai/benchmarks/skillsbench

- Category: [external](/external)
- Last updated: September 11, 2026

## About SkillsBench

- Year: 2026
- Tasks: Agent skill-importance tasks
- Format: Accuracy score
- Difficulty: Agent skill evaluation
- Paper: [SkillsBench](https://www.vals.ai/benchmarks/skillsbench)

BenchLM mirrors the public Vals AI SkillsBench leaderboard as display-only external evidence. The captured snapshot preserves overall scores, task-level scores where Vals publishes them, uncertainty, latency, and cost-per-test metadata. It is excluded from BenchLM weighted rankings.

SkillsBench is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (35 models)

| Rank | Model | Configuration | Creator | Score |
|------|-------|---------------|---------|-------|
| 1 | [DeepSeek V4.1 Flash](/models/deepseek-v4-1-flash) | OpenHands · high reasoning · OpenHands | DeepSeek | 69.80% |
| 2 | [Grok 4.5](/models/grok-4-5) | OpenHands · high reasoning · OpenHands | xAI | 66.03% |
| 3 | [Gemini 3.7 Flash](/models/gemini-3-7-flash) | OpenHands · high reasoning · OpenHands | Google | 65.89% |
| 4 | [GPT-5.5 Codex](https://www.vals.ai/models/openai_gpt-5.5-codex) | Codex · Codex | OpenAI | 62.55% |
| 5 | [GPT-5.5](/models/gpt-5-5) | OpenHands · xhigh reasoning · OpenHands | OpenAI | 62.21% |
| 6 | [Claude Fable 5.1](/models/claude-fable-5-1) | OpenHands · OpenHands | Anthropic | 61.55% |
| 7 | [GPT-5.6 Luna](/models/gpt-5-6-luna) | OpenHands · max reasoning · OpenHands | OpenAI | 60.45% |
| 8 | [Claude Opus 5](/models/claude-opus-5) | OpenHands · OpenHands | Anthropic | 60.44% |
| 9 | [Claude Opus 4.8](/models/claude-opus-4-8) | OpenHands · OpenHands | Anthropic | 59.23% |
| 10 | [Muse Spark 1.1](/models/muse-spark-1-1) | OpenHands · xhigh reasoning · OpenHands | Meta | 59.19% |
| 11 | [GPT-5.6 Terra](/models/gpt-5-6-terra) | OpenHands · max reasoning · OpenHands | OpenAI | 58.90% |
| 12 | [Gemini 3.8 Flash](/models/gemini-3-8-flash) | OpenHands · high reasoning · OpenHands | Google | 57.98% |
| 13 | [Grok 4.6](/models/grok-4-6) | OpenHands · high reasoning · OpenHands | xAI | 55.77% |
| 14 | [Qwen3.7 Plus](/models/qwen3-7-plus) | OpenHands · OpenHands | Alibaba | 54.30% |
| 15 | [GPT-5.6 Sol](/models/gpt-5-6-sol) | OpenHands · max reasoning · OpenHands | OpenAI | 54.10% |
| 16 | [DeepSeek V4 Pro 0813](/models/deepseek-v4-pro-0813) | OpenHands · max reasoning · OpenHands | DeepSeek | 53.83% |
| 17 | [Muse Spark 1.2](/models/muse-spark-1-2) | OpenHands · xhigh reasoning · OpenHands | Meta | 53.04% |
| 18 | [Gemini 3.5 Flash](/models/gemini-3-5-flash) | OpenHands · high reasoning · OpenHands | Google | 52.74% |
| 19 | [GPT-5.4](/models/gpt-5-4) | OpenHands · xhigh reasoning · OpenHands | OpenAI | 51.71% |
| 20 | [MiniMax M3](/models/minimax-m3) | OpenHands · OpenHands | MiniMax | 51.50% |
| 21 | [DeepSeek V4 Pro 0813](/models/deepseek-v4-pro-0813) | OpenHands · max reasoning · OpenHands | DeepSeek | 51.27% |
| 22 | [DeepSeek V4 Flash 0731](/models/deepseek-v4-flash-0731) | OpenHands · high reasoning · OpenHands | DeepSeek | 50.67% |
| 23 | [Kimi K2.7 Code](/models/kimi-k2-7-code) | OpenHands · OpenHands | Moonshot AI | 50.04% |
| 24 | [Claude Sonnet 4.6](/models/claude-sonnet-4-6) | OpenHands · OpenHands | Anthropic | 49.05% |
| 25 | [GLM-5.3](/models/glm-5-3) | OpenHands · max reasoning · OpenHands | Z.AI | 47.51% |
| 26 | [Claude Sonnet 5](/models/claude-sonnet-5) | OpenHands · OpenHands | Anthropic | 46.48% |
| 27 | [GLM-5.2](/models/glm-5-2) | OpenHands · max reasoning · OpenHands | Z.AI | 45.08% |
| 28 | [Qwen3.8 Max](/models/qwen3-8-max) | OpenHands · OpenHands | Alibaba | 42.01% |
| 29 | [Grok 4.3](/models/grok-4-3) | OpenHands · high reasoning · OpenHands | xAI | 40.64% |
| 30 | [GLM-5.3-Flash](/models/glm-5-3-flash) | OpenHands · max reasoning · OpenHands | Z.AI | 40.16% |
| 31 | [Qwen3.8-27B](/models/qwen3-8-27b) | OpenHands · xhigh reasoning · OpenHands | Alibaba | 38.11% |
| 32 | [Inkling-Small](/models/inkling-small) | OpenHands · 0.99 reasoning · OpenHands | Thinking Machines Lab | 33.58% |
| 33 | [Ling 3.0 Flash 2607](https://www.vals.ai/models/ant_ling-3.0-flash-2607) | OpenHands · OpenHands | Ant | 27.95% |
| 34 | [Inkling](/models/inkling) | OpenHands · 0.99 reasoning · OpenHands | Thinking Machines Lab | 26.09% |
| 35 | [Mercury 2.5](/models/mercury-2-5) | OpenHands · high reasoning · OpenHands | Inception | 18.06% |

## FAQ

### What does SkillsBench measure?

How important are skills for agents?

### Which model leads the published SkillsBench snapshot?

DeepSeek V4.1 Flash currently leads the published SkillsBench snapshot with a score of 69.80%.

### How many models are evaluated on SkillsBench?

The September 11, 2026 contains 35 AI models.

### Does SkillsBench affect BenchLM's overall score?

Not directly. SkillsBench is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.