llmrun Composite Score · Reasoning
The best open LLMs for reasoning, ranked on hard multi-step reasoning boards such as ARC-AGI and HLE. The score shown is the overall llmrun Composite Score; the highlighted column is the reasoning skill score used for the order.
45 benchmarks4 sourcesUpdated 17 Oct 2026Reference scale 2026-Q4How the Composite Score is built
Reasoning score vs VRAM needed
This chart doesn't show proprietary models.
| # (open / overall) | Model | llmrun Composite Score | Coding | Agents | Math | Science | Reasoning | Benchmarks | VRAM |
|---|---|---|---|---|---|---|---|---|---|
| 1 / 31 | DeepSeek V4.1 Flash458 GB · 24 benchmarks | 122 | 128 | 129 | 123 | 134 | 24 | 458 GB | |
| 2 / 38 | GLM 5.3454 GB · 26 benchmarks | 130 | 133 | 129 | 131 | 130 | 26 | 454 GB | |
| 3 / 42 | DeepSeek V4 Flash 0731183 GB · 21 benchmarks | 121 | – | 128 | 124 | 129 | 21 | 183 GB | |
| 4 / 46 | GLM 5.3 Flash205 GB · 24 benchmarks | 117 | 127 | 121 | 124 | 127 | 24 | 205 GB | |
| 5 / 57 | Smaug Flash208 GB · 6 benchmarks | — | 113 | – | – | – | 124 | 6 | 208 GB |
| 6 / 58 | Qwen3.8 27B17.9 GB · 15 benchmarks | 120 | 123 | 118 | 112 | 124 | 15 | 17.9 GB | |
| 7 / 61 | DeepSeek V4 Flash Vision Exp183 GB · 6 benchmarks | — | 101 | – | – | – | 123 | 6 | 183 GB |
| 8 / 64 | Inkling Small163 GB · 11 benchmarks | – | – | 117 | 118 | 123 | 11 | 163 GB | |
| 9 / 72 | GLM 5.2454 GB · 26 benchmarks | 120 | 128 | 126 | 129 | 120 | 26 | 454 GB | |
| 10 / 79 | Qwen3.5 397B A17B243 GB · 8 benchmarks | – | – | 112 | – | 117 | 8 | 243 GB | |
| 11 / 81 | NVIDIA Nemotron 3 Ultra 550B A55B BF16378 GB · 17 benchmarks | 104 | 102 | 112 | 112 | 116 | 17 | 378 GB | |
| 12 / 82 | GLM 5.1454 GB · 14 benchmarks | 115 | 123 | 117 | 117 | 116 | 14 | 454 GB | |
| 13 / 91 | MiniMax M2.5140 GB · 11 benchmarks | 109 | 110 | – | – | 114 | 11 | 140 GB | |
| 14 / 92 | DeepSeek V3.2 Exp413 GB · 10 benchmarks | 107 | – | – | 101 | 113 | 10 | 413 GB | |
| 15 / 94 | Qwen3.5 122B A10B75.5 GB · 6 benchmarks | – | – | – | 96 | 112 | 6 | 75.5 GB | |
| 16 / 97 | DeepSeek V3.2413 GB · 13 benchmarks | – | 108 | – | – | 112 | 13 | 413 GB | |
| 17 / 98 | DeepSeek V4 Flash175 GB · 10 benchmarks | 107 | – | – | 111 | 112 | 10 | 175 GB | |
| 18 / 99 | GLM 4.7218 GB · 13 benchmarks | 101 | 106 | 111 | 111 | 111 | 13 | 218 GB | |
| 19 / 103 | Qwen3.5 27B17.9 GB · 6 benchmarks | 98 | – | – | – | 111 | 6 | 17.9 GB | |
| 20 / 104 | Gemma 4 31B IT21.7 GB · 13 benchmarks | 108 | – | – | 105 | 110 | 13 | 21.7 GB | |
| 21 / 109 | GLM 5454 GB · 14 benchmarks | 110 | 118 | – | – | 109 | 14 | 454 GB | |
| 22 / 112 | MiniMax M3258 GB · 22 benchmarks | 115 | 112 | 106 | 118 | 108 | 22 | 258 GB | |
| 23 / 113 | DeepSeek V3.1 Terminus412 GB · 5 benchmarks | 107 | – | – | 99 | 108 | 5 | 412 GB | |
| 24 / 116 | Qwen3.5 35B A3B22.0 GB · 7 benchmarks | – | – | – | 108 | 107 | 7 | 22.0 GB | |
| 25 / 117 | Qwen3 235B A22B Thinking 2507143 GB · 12 benchmarks | 102 | – | – | 108 | 106 | 12 | 143 GB | |
| 26 / 122 | DeepSeek V3.1412 GB · 7 benchmarks | – | – | – | – | 105 | 7 | 412 GB | |
| 27 / 123 | Qwen3.6 27B17.9 GB · 11 benchmarks | 102 | – | 114 | 111 | 105 | 11 | 17.9 GB | |
| 28 / 125 | Qwen3.6 35B A3B22.0 GB · 11 benchmarks | 105 | – | 109 | 110 | 104 | 11 | 22.0 GB | |
| 29 / 130 | Gemma 4 26B A4B IT16.5 GB · 8 benchmarks | 104 | – | – | 102 | 102 | 8 | 16.5 GB | |
| 30 / 133 | DeepSeek R1 0528412 GB · 10 benchmarks | 107 | – | – | – | 100 | 10 | 412 GB | |
| 31 / 135 | Qwen3 235B A22B143 GB · 9 benchmarks | 101 | – | – | – | 99 | 9 | 143 GB | |
| 32 / 137 | Qwen3 235B A22B Instruct 2507143 GB · 8 benchmarks | 102 | – | – | – | 98 | 8 | 143 GB | |
| 33 / 139 | Qwen3.5 9B6.6 GB · 8 benchmarks | – | – | – | 105 | 97 | 8 | 6.6 GB | |
| 34 / 141 | GPT OSS 120B70.8 GB · 19 benchmarks | 98 | 82 | – | 103 | 96 | 19 | 70.8 GB | |
| 35 / 142 | DeepSeek R1412 GB · 13 benchmarks | 100 | – | – | 99 | 96 | 13 | 412 GB | |
| 36 / 144 | Mistral Small 4 119B 260381.6 GB · 3 benchmarks | – | – | – | – | 96 | 3 | 81.6 GB | |
| 37 / 147 | Qwen3 30B A3B Thinking 250719.4 GB · 7 benchmarks | – | – | – | 99 | 94 | 7 | 19.4 GB | |
| 38 / 149 | Qwen3 30B A3B Instruct 250719.4 GB · 4 benchmarks | – | – | – | – | 93 | 4 | 19.4 GB | |
| 39 / 154 | Qwen3 32B22.6 GB · 8 benchmarks | 93 | – | – | 96 | 92 | 8 | 22.6 GB | |
| 40 / 155 | GPT OSS 20B13.1 GB · 10 benchmarks | 101 | – | – | 92 | 91 | 10 | 13.1 GB | |
| 41 / 161 | Mistral Large 3 675B Instruct 2512454 GB · 5 benchmarks | 93 | – | – | 97 | 89 | 5 | 454 GB | |
| 42 / 162 | Qwen3 14B11.9 GB · 7 benchmarks | – | – | – | 94 | 88 | 7 | 11.9 GB | |
| 43 / 163 | Magistral Small 250617.8 GB · 4 benchmarks | – | – | – | – | 88 | 4 | 17.8 GB | |
| 44 / 165 | DeepSeek v3 0324412 GB · 12 benchmarks | 100 | – | – | 98 | 88 | 12 | 412 GB | |
| 45 / 173 | Qwen2.5 72B Instruct49.3 GB · 8 benchmarks | – | – | – | – | 86 | 8 | 49.3 GB | |
| 46 / 174 | Llama 4 Maverick 17B 128E Instruct273 GB · 16 benchmarks | 81 | – | – | 96 | 85 | 16 | 273 GB | |
| 47 / 175 | Llama 3.1 405B Instruct276 GB · 7 benchmarks | – | – | – | – | 85 | 7 | 276 GB | |
| 48 / 178 | Mistral Large Instruct 241179.8 GB · 5 benchmarks | – | – | – | – | 84 | 5 | 79.8 GB | |
| 49 / 179 | Qwen3 30B A3B19.4 GB · 6 benchmarks | – | – | – | – | 84 | 6 | 19.4 GB | |
| 50 / 181 | C4ai Command A 03 202578.7 GB · 3 benchmarks | – | – | – | – | 83 | 3 | 78.7 GB | |
| 51 / 182 | Llama 3.1 70B Instruct51.3 GB · 7 benchmarks | – | – | – | – | 83 | 7 | 51.3 GB | |
| 52 / 183 | Mistral Large Instruct 240786.4 GB · 6 benchmarks | – | – | – | – | 83 | 6 | 86.4 GB | |
| 53 / 185 | Llama 3.3 70B Instruct51.3 GB · 12 benchmarks | 76 | – | – | 81 | 82 | 12 | 51.3 GB | |
| 54 / 186 | Qwen3 8B7.6 GB · 7 benchmarks | – | – | – | 88 | 81 | 7 | 7.6 GB | |
| 55 / 187 | DeepSeek v3412 GB · 8 benchmarks | 97 | – | – | 90 | 81 | 8 | 412 GB | |
| 56 / 188 | Llama 4 Scout 17B 16E Instruct77.0 GB · 11 benchmarks | – | – | – | 83 | 80 | 11 | 77.0 GB | |
| 57 / 191 | Mixtral 8x22B Instruct v0.188.4 GB · 5 benchmarks | – | – | – | – | 76 | 5 | 88.4 GB | |
| 58 / 192 | Gemma 3 27B IT21.7 GB · 9 benchmarks | 70 | – | – | 81 | 74 | 9 | 21.7 GB | |
| 59 / 193 | C4ai Command R Plus 08 202473.8 GB · 3 benchmarks | — | – | – | – | – | 74 | 3 | 73.8 GB |
| 60 / 197 | Llama 3.1 8B Instruct7.9 GB · 9 benchmarks | 56 | – | – | 56 | 69 | 9 | 7.9 GB | |
| 61 / 198 | Mixtral 8x7B Instruct v0.130.5 GB · 4 benchmarks | – | – | – | – | 69 | 4 | 30.5 GB | |
| 62 / 199 | Gemma 3 4B IT5.0 GB · 5 benchmarks | – | – | – | – | 69 | 5 | 5.0 GB | |
| 63 / 202 | Gemma 3 12B IT10.9 GB · 7 benchmarks | – | – | – | 71 | 66 | 7 | 10.9 GB | |
| 64 / 204 | Qwen2.5 7B Instruct5.8 GB · 5 benchmarks | – | – | – | – | 65 | 5 | 5.8 GB | |
| 65 / 206 | C4ai Command R 08 202425.1 GB · 2 benchmarks | — | – | – | – | – | 64 | 2 | 25.1 GB |
| 66 / 207 | Meta Llama 3 8B Instruct7.9 GB · 5 benchmarks | – | – | – | – | 64 | 5 | 7.9 GB | |
| 67 / 209 | Llama 2 70B Chat HF50.2 GB · 4 benchmarks | – | – | – | – | 45 | 4 | 50.2 GB |
The range after each score is a 90% interval: given the boards a model has results on, its true score is very likely inside it. Short ranges mean many agreeing results; long ranges mean few or conflicting ones.
Why some models are missing. A model receives a score only if it has results on at least 4 benchmarks across at least 2 categories, with one being reasoning or coding, or on 3 benchmarks that narrow its 90% range to at most 15.6 points. The results must come from at least two different benchmarks: categories of one benchmark, such as LiveBench, count once. Skill scores require at least 2 benchmarks for that skill; otherwise they display as –.
VRAM is llmrun's estimate at Q4_K_M (or the smallest quantization we track) for the weights plus room for a 16K-token context. Longer contexts need more memory. Proprietary models can't run locally, so the VRAM filter hides them.
How the llmrun Composite Score is built · llmrun doesn't run these benchmarks; scores come from public sources.