Coding
ALE-Bench Leaderboard
ALE-Bench, from Sakana AI and AtCoder, tests long-horizon algorithmic optimisation using problems drawn from AtCoder Heuristic Contests, where there is no single correct answer and solutions are scored on quality. Results are reported as an AtCoder-style performance rating, where higher is better.
Source: epoch51 open models ranked+87 proprietaryData through Nov 2026
Open models ranked on ALE-Bench
# shows rank among open models / rank overall (including proprietary).
| # | Model | Score | VRAM @ Q4 |
|---|---|---|---|
| 1 / 14 | Kimi K3.1 Pro · 2780.1B | 1932.8 | ~1674 GB |
| 2 / 21 | DeepSeek V4.1 Flash 1017 · 763.2B | 1622.0 | ~458 GB |
| 3 / 23 | Muse Spark Open 816B A42B · 822.0B | 1590.1 | ~543 GB |
| 4 / 29 | Kimi K3 · 2779.9B | 1524.5 | ~1674 GB |
| 5 / 34 | DeepSeek V4 Pro 0813 · 1650.5B | 1403.2 | ~991 GB |
| 6 / 38 | GLM 5.3 · 753.3B | 1317.4 | ~456 GB |
| 7 / 40 | DeepSeek V4 Flash 0731 · 304.2B | 1306.1 | ~183 GB |
| 8 / 46 | MiMo V2.6 Flash RL · 309B | 1244.2 | ~186 GB |
| 9 / 48 | Mistral Large 4 1T A52B · 1000B | 1213.2 | ~604 GB |
| 10 / 56 | MiMo V2.6 Pro RL · 1020B | 1158.1 | ~613 GB |
| 11 / 59 | Beam 1.0 501B A23B · 501.3B | 1097.5 | ~301 GB |
| 12 / 60 | Kimi K2.6 · 1026.9B | 1092.7 | ~620 GB |
| 13 / 61 | DeepSeek V4.1 Flash · 763.2B | 1092.3 | ~458 GB |
| 14 / 64 | GLM 5.2 · 753.3B | 1047.0 | ~456 GB |
| 15 / 66 | DeepSeek V4 Pro · 1598.8B | 1006.1 | ~960 GB |
| 16 / 70 | Inkling · 952.4B | 946.0 | ~572 GB |
| 17 / 73 | Gemma 4 26B A4B IT · 25.8B | 927.2 | ~16 GB |
| 18 / 74 | Gemma 4 31B IT · 31.3B | 925.5 | ~20 GB |
| 19 / 77 | MiMo V2.5 Pro · 1023.2B | 899.8 | ~614 GB |
| 20 / 78 | GLM 5.1 · 753.9B | 887.1 | ~457 GB |
| 21 / 79 | Kimi K2.7 Code · 1026.9B | 886.2 | ~620 GB |
| 22 / 81 | Kimi K2.5 · 1026.9B | 821.6 | ~620 GB |
| 23 / 83 | DeepSeek R1 0528 · 684.5B | 804.1 | ~415 GB |
| 24 / 90 | GLM 5 · 753.9B | 765.6 | ~457 GB |
| 25 / 92 | Mistral Medium 3.5 128B · 127.7B | 764.0 | ~78 GB |
| 26 / 93 | DeepSeek V3.1 Terminus · 684.5B | 745.2 | ~415 GB |
| 27 / 94 | MiMo v2 Flash · 309.8B | 738.0 | ~186 GB |
| 28 / 97 | Step 3.7 Flash · 201.4B | 694.1 | ~133 GB |
| 29 / 98 | DeepSeek V4 Flash · 290.9B | 678.2 | ~175 GB |
| 30 / 104 | MiniMax M3 · 427.0B | 640.0 | ~257 GB |
| 31 / 105 | MiniMax M2.1 · 228.7B | 623.8 | ~138 GB |
| 32 / 106 | Qwen3.5 397B A17B · 403.4B | 621.9 | ~242 GB |
| 33 / 107 | MiniMax M2.5 · 228.7B | 618.2 | ~138 GB |
| 34 / 109 | MiniMax M2.7 · 228.7B | 599.3 | ~138 GB |
| 35 / 110 | Kimi K2 Thinking · 1026.4B | 597.5 | ~620 GB |
| 36 / 112 | GPT OSS 120B · 116.8B | 575.6 | ~71 GB |
| 37 / 113 | GPT OSS 20B · 20.9B | 566.0 | ~13 GB |
| 38 / 115 | MiMo V2.5 · 310.8B | 514.0 | ~187 GB |
| 39 / 116 | Mistral Small 4 119B 2603 · 119.4B | 497.6 | ~73 GB |
| 40 / 117 | Qwen3 Coder 480B A35B Instruct · 480.2B | 461.4 | ~289 GB |
| 41 / 119 | Ring 2.6 1T · 1025.4B | 432.6 | ~2057 GB |
| 42 / 120 | GLM 4.7 · 358.3B | 399.5 | ~216 GB |
| 43 / 123 | Qwen3.5 27B · 27.8B | 349.4 | ~17 GB |
| 44 / 124 | GLM 4.5 · 358.3B | 344.8 | ~216 GB |
| 45 / 125 | GLM 4.6 · 356.8B | 340.8 | ~215 GB |
| 46 / 128 | GLM 5.3 Flash · 321.3B | 303.6 | ~195 GB |
| 47 / 130 | Kimi K2 Instruct 0905 · 1026.4B | 267.1 | ~620 GB |
| 48 / 132 | Mistral Large 3 675B Instruct 2512 · 675B | 264.7 | ~446 GB |
| 49 / 134 | Qwen3.5 35B A3B · 36.0B | 221.8 | ~22 GB |
| 50 / 135 | NVIDIA Nemotron 3 Super 120B A12B BF16 · 123.6B | 213.9 | ~75 GB |
| 51 / 137 | Llama 4 Maverick 17B 128E Instruct · 401.6B | 173.0 | ~265 GB |
Score vs model size
Which models give the most quality for their size — the ones worth running locally.
- GPT OSS 20B, 21B, score 566.0 — on the efficiency frontier (best score at its size or smaller).
- Gemma 4 26B A4B IT, 26B, score 927.2 — on the efficiency frontier (best score at its size or smaller).
- DeepSeek V4 Flash 0731, 304B, score 1306.1 — on the efficiency frontier (best score at its size or smaller).
- GLM 5.3, 753B, score 1317.4 — on the efficiency frontier (best score at its size or smaller).
- DeepSeek V4.1 Flash 1017, 763B, score 1622.0 — on the efficiency frontier (best score at its size or smaller).
- Kimi K3.1 Pro, 2.8T, score 1932.8 — on the efficiency frontier (best score at its size or smaller).
ALE-Bench: frequently asked questions
- What is the best open LLM on ALE-Bench?
- Kimi K3.1 Pro is the top open model on ALE-Bench, scoring 1932.8. Among all models tested — including proprietary ones — it ranks #14. The top model overall is GPT 6 Astra Max (OpenAI) at 2951.3.
- What's the best ALE-Bench model you can run on a 24 GB GPU?
- Gemma 4 26B A4B IT is the highest-scoring open model that fits in 24 GB at 4-bit quantization (~16 GB), scoring 927.2 on ALE-Bench.
- Can open models match proprietary models on ALE-Bench?
- Not quite on ALE-Bench: the strongest proprietary model (GPT 6 Astra Max) scores 2951.3, ahead of the best open model (Kimi K3.1 Pro) at 1932.8 — but you can run the open one yourself.
Scores aggregated from epoch. llmrun does not run this benchmark — see the source for methodology, or the about benchmarks for what it measures.