LLM Stats (Program Bench): leaderboard
Program Bench as reported by LLM Stats — program-understanding and program-generation evaluation; leaderboard score.
Metric: Score (%). Source: llm-stats.com. Saturation forecast: Not forecast. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 | 91.2 | #3 |
| 2 | Claude Haiku 5.5 | 82 | #40 |
| 3 | Claude Sonnet 5.5 | 79.7 | #11 |
| 4 | Kimi K3 | 77.8 | #30 |
| 5 | GLM-5.2 | 63.7 | #73 |
| 6 | Muse Wonder Full | 59 | #10 |
| 7 | Kimi K2.7 Code | 53.6 | #89 |
| 8 | Seed 2.1 Pro | 50.3 | #67 |
| 9 | Muse Wonder | 50 | #15 |
| 10 | Seed 2.1 Turbo | 49.4 | #86 |
| 11 | DeepSeek V4.1 Flash (1017) | 29.1 | #20 |
| 12 | MiMo-V2.6-Pro | 26.5 | #42 |
| 13 | MiMo-V2.6-Flash | 26 | #70 |
| 14 | DeepSeek V4.1 Flash | 20.3 | #36 |
| 15 | GLM-5.3 | 19 | #39 |
No result here: #6 GPT-6 Astra, #7 GPT-6.1 Sol, #8 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-program-bench · How It Works · Data refreshed daily, snapshot 2026-10-20.