Skip to content

Benchmark

FrontierSWE

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

Models measured
16
Scores recorded
22
Sources
2
IQ weight
1.33

Score distribution

62.3%
Mean
73.5%
Median
21.1%
Std dev
27%
10th percentile
81.2%
90th percentile
22%
Min
22%90%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.33 for chat models, measured over 16 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 5· LLM Stats data90%
2Claude Fable 5Qwen表:Claude Code;对照为 2026-08-03 官方榜 MEAN@5· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.88.8%
3Kimi K3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim81.2%
4Kimi K3Kimi Code · Dominance · snapshot 2026-07-16· Reported by Moonshot AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.81.2%
5GLM 5.3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim78.1%
6GLM 5.3Dominance · Proximal · 1M · max effort · 128K output · snapshot 2026-08-14· Reported by Z.ai· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.78.1%
7Claude Opus 4.8· LLM Stats data75%
8GLM 5.2Dominance · 1M context · max effort · 128K output · as of 2026-06-16· Reported by Z.ai· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.74.4%
9GLM 5.2· LLM Stats data74%
10Qwen3.8 Max· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim73.5%
11Qwen3.8 MaxQwen表:Claude Code;对照为 2026-08-03 官方榜 MEAN@5· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.73.5%
12GPT 5.5· LLM Stats data73%
13Claude Opus 4.8Qwen表:Claude Code;对照为 2026-08-03 官方榜 MEAN@5· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.70%
14Claude Opus 4.7· LLM Stats data63%
15Claude Opus 4.6· LLM Stats data56%
16GPT 5.4· LLM Stats data54%
17Qwen3.7 MaxQwen表:Claude Code;对照为 2026-08-03 官方榜 MEAN@5· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.40.7%
18Gemini 3.1 Pro Preview· LLM Stats data40%
19GLM 5.1· LLM Stats data31%
20Kimi K2.6· LLM Stats data27%
21Kimi K2.5· LLM Stats data26%
22Qwen3.6 Plus· LLM Stats data22%

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