AI Coding Daily (OpenCode) - Total: leaderboard
AI Coding Daily's LLM Coding Leaderboard (Povilas Korop) tests each model and reasoning setting on two code-quality projects (Laravel, React-TS; 20 points each), a bug-finding test (10 points), and four edge-case reliability projects (5 points each). This board is total points (currently out of 70) for runs in the OpenCode harness; higher is better.
Metric: Total points (max 70). Source: aicodingdaily.com. Saturation forecast: Around February 2027. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K3.1 Pro (Max) | 63.8 | #17 (Kimi K3.1 Pro) |
| 2 | DeepSeek V4.1 Flash (1017) (Max) | 60.47 | #27 (DeepSeek V4.1 Flash (1017)) |
| 3 | Kimi K3.1 Pro (High) | 58.46 | #17 (Kimi K3.1 Pro) |
| 4 | DeepSeek V4.1 Flash (1017) (High) | 54.71 | #27 (DeepSeek V4.1 Flash (1017)) |
| 5 | GLM-5.3 (High) | 54.32 | #44 (GLM-5.3) |
| 6 | MiniMax-M3.1-Flash (Max) | 53.59 | #52 (MiniMax-M3.1-Flash) |
| 7 | Muse Spark 1.3 (Max) | 52.51 | #31 (Muse Spark 1.3) |
| 8 | Qwen 3.8 Flash (Max) | 52.14 | #62 (Qwen 3.8 Flash) |
| 9 | mistral-large-4-oss (High) | 50.31 | #60 (mistral-large-4-oss) |
| 10 | Hy3 (High) | 49.37 | #100 (Hy3) |
| 11 | DeepSeek V4 Pro (0813) (Max) | 49.35 | #58 (DeepSeek V4 Pro (0813)) |
| 12 | DeepSeek V4 Pro (0813) (High) | 47.34 | #58 (DeepSeek V4 Pro (0813)) |
| 13 | Qwen 3.8 27B (xHigh) | 45.26 | #123 (Qwen 3.8 27B) |
| 14 | GLM-5.3 Flash (Max) | 45.01 | #77 (GLM-5.3 Flash) |
| 15 | Gemini 3.1 Pro (Preview) | 41.36 | #64 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=ai-coding-daily-opencode-total · How It Works · Data refreshed daily, snapshot 2026-11-05.