Benchmark
DeepSWE
Coding benchmarks compare models on programming tasks. Scores are percentages.
All coding benchmarksScores from DeepSWE, LLM Stats data.
- Models measured
- 36
- Scores recorded
- 36
- Sources
- 2
- IQ weight
- 2.49
Score distribution
- 55.3%
- Mean
- 59.3%
- Median
- 16.5%
- Std dev
- 30.5%
- 10th percentile
- 72.8%
- 90th percentile
- 1.6%
- Min
This benchmark has a weight of 2.49 for chat models, measured over 36 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | |
|---|---|---|
| 1 | Kimi K3.1 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 75.5% |
| 2 | GPT-6 Astra· DeepSWE | 74.1% |
| 3 | Gemini 3.8 Flash· DeepSWE | 73.8% |
| 4 | Claude Opus 5· DeepSWE | 73.7% |
| 5 | GPT-5.6 Sol· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 72.7% |
| 6 | Claude Fable 5· DeepSWE | 69.9% |
| 7 | GPT-5.6 Terra· DeepSWE | 69.6% |
| 8 | GLM 5.3· DeepSWE | 69% |
| 9 | Kimi K3· DeepSWE | 68.5% |
| 10 | Grok 4.6· DeepSWE | 67.5% |
| 11 | GPT-5.6 Luna· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 67.2% |
| 12 | GPT 5.5· DeepSWE | 67% |
| 13 | Gemini 3.7 Flash· DeepSWE | 65.5% |
| 14 | Hy4 preview· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 64.3% |
| 15 | GLM 5.3 Flash· DeepSWE | 63.4% |
| 16 | DeepSeek V4 Pro· DeepSWE | 62.8% |
| 17 | DeepSeek V4 Pro 0813· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 62.7% |
| 18 | DeepSeek V4 Flash Vision Exp· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 59.3% |
| 19 | Claude Opus 4.8· DeepSWE | 59% |
| 20 | Qwen3.8 Max· DeepSWE | 57.5% |
| 21 | Muse Spark 1.2· DeepSWE | 54.9% |
| 22 | DeepSeek V4 Flash 0731· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 54.4% |
| 23 | Claude Sonnet 5· DeepSWE | 53.9% |
| 24 | Grok 4.5· DeepSWE | 53.8% |
| 25 | DeepSeek V4 Flash· DeepSWE | 53.3% |
| 26 | Muse Spark 1.1· DeepSWE | 53.3% |
| 27 | GPT 5.4· DeepSWE | 51.8% |
| 28 | Gemini 3.6 Flash· DeepSWE | 46.7% |
| 29 | GLM 5.2· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 46.2% |
| 30 | Gemini 3.5 Flash· DeepSWE | 36.1% |
| 31 | Doubao Seed 2.1 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 32.7% |
| 32 | Kimi K2.7 Code· DeepSWE | 30.5% |
| 33 | Claude Sonnet 4.6· DeepSWE | 29.9% |
| 34 | Hy3 (Deep Infra)· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 28% |
| 35 | Doubao Seed 2.1 Turbo· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 23% |
| 36 | Gemini 3.1 Pro Preview· DeepSWE | 11.7% |
Tokens the model generated per task, as measured by the source. Comparable within this benchmark only.
Data credits: Data by LLM Stats