Skip to main content
BenchLM
Data

Vals SkillsBench (SkillsBench)

We show this table for reference; we do not rank on it.

Data verified

How important are skills for agents?

Evaluation results published by Vals AI. Published leaderboard aggregates and evaluation settings only. Dataset contents are not included.

SkillsBench score on SkillsBench — September 27, 2026

We mirror the published skillsbench score view for SkillsBench. DeepSeek V4.1 Flash 1017 leads the public snapshot at 69.80%, followed by Grok 4.5 (66.03%) and Gemini 3.7 Flash (65.89%). We do not use these results to rank models overall.

1Open weights

DeepSeek V4.1 Flash 1017

DeepSeek

OpenHands · high reasoning

OpenHands

69.80%
Overall 65.22Context 1M
2Closed

Grok 4.5

xAI

OpenHands · high reasoning

OpenHands

66.03%
Overall 62.92Context 500K
3Closed

Gemini 3.7 Flash

Google

OpenHands · high reasoning

OpenHands

65.89%
Overall 66.53Context 1M

35 modelsAgenticCurrentDisplay onlyUpdated September 27, 2026

SkillsBench score table (35 models)

SkillsBench score results for SkillsBench
RankModel / configurationScoreParameters (B)Open / closed
1DeepSeek V4.1 Flash 1017DeepSeekOpenHands · high reasoningOpenHands
69.80%
Not reportedOpen
2Grok 4.5xAIOpenHands · high reasoningOpenHands
66.03%
Not reportedClosed
3Gemini 3.7 FlashGoogleOpenHands · high reasoningOpenHands
65.89%
Not reportedClosed
4GPT-5.5 CodexOpenAICodexCodex
62.55%
Not reportedUnknown
5GPT-5.5OpenAIOpenHands · xhigh reasoningOpenHands
62.21%
Not reportedClosed
6Claude Fable 5.1AnthropicOpenHandsOpenHands
61.55%
Not reportedClosed
7GPT-5.6 LunaOpenAIOpenHands · max reasoningOpenHands
60.45%
Not reportedClosed
8Claude Opus 5AnthropicOpenHandsOpenHands
60.44%
Not reportedClosed
9Claude Opus 4.8AnthropicOpenHandsOpenHands
59.23%
Not reportedClosed
10Muse Spark 1.1MetaOpenHands · xhigh reasoningOpenHands
59.19%
Not reportedClosed
11GPT-5.6 TerraOpenAIOpenHands · max reasoningOpenHands
58.90%
Not reportedClosed
12Gemini 3.8 FlashGoogleOpenHands · high reasoningOpenHands
57.98%
Not reportedClosed
13Grok 4.6xAIOpenHands · high reasoningOpenHands
55.77%
Not reportedClosed
14Qwen3.7 PlusAlibabaOpenHandsOpenHands
54.30%
Not reportedClosed
15GPT-5.6 SolOpenAIOpenHands · max reasoningOpenHands
54.10%
Not reportedClosed
16DeepSeek V4 Pro 0813DeepSeekOpenHands · max reasoningOpenHands
53.83%
Not reportedOpen
17Muse Spark 1.2MetaOpenHands · xhigh reasoningOpenHands
53.04%
Not reportedClosed
18Gemini 3.5 FlashGoogleOpenHands · high reasoningOpenHands
52.74%
Not reportedClosed
19GPT-5.4OpenAIOpenHands · xhigh reasoningOpenHands
51.71%
Not reportedClosed
20MiniMax M3MiniMaxOpenHandsOpenHands
51.50%
Not reportedOpen
21DeepSeek V4 Pro 0813DeepSeekOpenHands · max reasoningOpenHands
51.27%
Not reportedOpen
22DeepSeek V4 Flash 0731DeepSeekOpenHands · high reasoningOpenHands
50.67%
Not reportedOpen
23Kimi K2.7 CodeMoonshot AIOpenHandsOpenHands
50.04%
Not reportedOpen
24Claude Sonnet 4.6AnthropicOpenHandsOpenHands
49.05%
Not reportedClosed
25GLM-5.3Z.AIOpenHands · max reasoningOpenHands
47.51%
Not reportedOpen
26Claude Sonnet 5AnthropicOpenHandsOpenHands
46.48%
Not reportedClosed
27GLM-5.2Z.AIOpenHands · max reasoningOpenHands
45.08%
Not reportedOpen
28Qwen3.8 MaxAlibabaOpenHandsOpenHands
42.01%
Not reportedOpen
29Grok 4.3xAIOpenHands · high reasoningOpenHands
40.64%
Not reportedClosed
30GLM-5.3-FlashZ.AIOpenHands · max reasoningOpenHands
40.16%
Not reportedOpen
31Qwen3.8-27BAlibabaOpenHands · xhigh reasoningOpenHands
38.11%
Not reportedOpen
32Inkling-SmallThinking Machines LabOpenHands · 0.99 reasoningOpenHands
33.58%
Not reportedOpen
33Ling 3.0 Flash 2607AntOpenHandsOpenHands
27.95%
Not reportedUnknown
34InklingThinking Machines LabOpenHands · 0.99 reasoningOpenHands
26.09%
Not reportedOpen
35Mercury 2.5InceptionOpenHands · high reasoningOpenHands
18.06%
Not reportedClosed

How SkillsBench is shown here

BenchLM mirrors the public Vals AI SkillsBench leaderboard captured from https://www.vals.ai/benchmarks/skillsbench and updated by Vals on September 27, 2026. The snapshot preserves overall scores, uncertainty, latency, cost-per-test metadata, and task-level scores where Vals publishes them.

SkillsBench is display only on BenchLM. Vals proprietary or Vals-hosted aggregate views are useful context, but BenchLM does not use them as weighted ranking inputs or as a replacement for benchmark-native source records.

Snapshot

35 Vals rows3 task viewspublic datasetTasks: Overall, No Skills, With SkillsDisplay only
SkillsBench on Vals AIVals methodologyVals home

The published SkillsBench snapshot places DeepSeek V4.1 Flash 1017 first at 69.80%. The third row is 3.91 points behind. The broader top-10 range is 10.61 points, so the table still separates the published systems.

35 models have been evaluated on SkillsBench. The benchmark falls in the Agentic category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. SkillsBench is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About SkillsBench

Year

2026

Tasks

Agent skill-importance tasks

Format

Accuracy score

Difficulty

Agent skill evaluation

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.

SkillsBenchPublic benchmark source

Freshness and provenance

Version

SkillsBench 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Questions

What does SkillsBench measure?

How important are skills for agents?

Which model leads the published SkillsBench snapshot?

DeepSeek V4.1 Flash 1017 currently leads the published SkillsBench snapshot with 69.80% skillsbench score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on SkillsBench?

The September 27, 2026 snapshot contains 35 AI models.

Last updated: September 27, 2026 · mirrored from the public benchmark leaderboard

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 6,100+ subscribers.

One email each week. Unsubscribe anytime.