Skip to content

Benchmark

MLS-Bench-Lite

Coding benchmarks compare models on programming tasks. Scores are percentages. Higher is better.

Models measured
8
Scores recorded
11
Sources
2
IQ weight
0.81

Score distribution

40.6%
Mean
41%
Median
7.2%
Std dev
31.7%
10th percentile
48.3%
90th percentile
26.7%
Min
26.7%49.9%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.81 for chat models, measured over 8 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Claude Fable 5Qwen表:Claude Code · 5h · max_tokens 131072;对照来自官方榜· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.49.9%
2Kimi K3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim48.3%
3Kimi K3max effort · temp 1.0 · top_p 1.0· Reported by Moonshot AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.48.3%
4GPT-5.6 SolQwen表:Claude Code · 5h · max_tokens 131072;对照来自官方榜· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.46.2%
5Claude Opus 4.8Qwen表:Claude Code · 5h · max_tokens 131072;对照来自官方榜· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.42.8%
6Qwen3.8 Max· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim41%
7Qwen3.8 MaxQwen表:Claude Code · 5h · max_tokens 131072;对照来自官方榜· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.41%
8Kimi K2.7 Code· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim35.1%
9Kimi K2.7 CodeKimi K2.7 Code official table · Kimi Code CLI · thinking + preserve_thinking · temp 1.0 · top_p .95 · 262,144 context · official 30-task subset · five-hour agent budget· Reported by Moonshot AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.35.1%
10Qwen3.7 MaxQwen表:Claude Code · 5h · max_tokens 131072;对照来自官方榜· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.31.7%
11Kimi K2.6Kimi K2.7 Code official comparison table · Kimi K2.6 · Kimi Code CLI · thinking · temp 1.0 · top_p .95 · 262,144 context · official 30-task subset · five-hour agent budget· Reported by Moonshot AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.26.7%

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · Data by LLM Stats