llmrun

llmrun Composite Score

The best LLMs, open and proprietary together, on one open-ended scale where 100 is the average model (overall capability). Switch to open models to see only what you can download and run.

46 benchmarks4 sourcesUpdated 14 Nov 2026Reference scale 2026-Q4How the Composite Score is built

Composite Score vs VRAM needed

This chart doesn't show proprietary models.

Models ranked by overall llmrun Composite Score
#Modelllmrun Composite ScoreCodingAgentsMathScienceReasoningBenchmarksVRAM
121Grok Build 0.15 benchmarksCoding 111Agents –Math –Science 118Reasoning –5–
122Qwen3 235B A22B Thinking 2507open143 GB · 13 benchmarksCoding 102Agents –Math –Science 108Reasoning 10813143 GB
123Gemini 2.5 Pro16 benchmarksCoding 112Agents 102Math 109Science 113Reasoning 10816–
124MiMo V2.5open188 GB · 4 benchmarksCoding 103Agents –Math –Science 107Reasoning –4188 GB
125Kimi K2 Thinkingopen617 GB · 9 benchmarksCoding 103Agents 108Math –Science –Reasoning –9617 GB
126MiniMax M2.7open140 GB · 7 benchmarksCoding 103Agents 112Math –Science 112Reasoning –7140 GB
127DeepSeek V3.1 Terminusopen412 GB · 5 benchmarksCoding 107Agents –Math –Science 99Reasoning 1085412 GB
128GPT 5.5 Instant6 benchmarksCoding –Agents –Math 105Science 112Reasoning –6–
129Qwen3.6 35B A3Bopen22.0 GB · 11 benchmarksCoding 105Agents –Math 109Science 110Reasoning 1041122.0 GB
130Claude Opus 4.1 (Aug 05, 2025)14 benchmarksCoding 106Agents 108Math 103Science 105Reasoning 11714–
131Gemini 2.5 Pro Preview (Mar 25)5 benchmarksCoding –Agents –Math –Science –Reasoning 1065–
132Gemma 4 31B ITopen21.7 GB · 14 benchmarksCoding 107Agents –Math –Science 105Reasoning 1111421.7 GB
133Qwen3.5 Flash10 benchmarksCoding –Agents –Math 106Science –Reasoning 11010–
134Gemma 4 26B A4B ITopen16.5 GB · 8 benchmarksCoding 104Agents –Math –Science 102Reasoning 102816.5 GB
135Qwen3.6 Flash10 benchmarksCoding –Agents –Math 108Science –Reasoning 10310–
136Claude Opus 4 (May 14, 2025)13 benchmarksCoding 107Agents –Math –Science 104Reasoning 10813–
137Mercury 24 benchmarksCoding 106Agents –Math –Science 100Reasoning –4–
138Claude Haiku 4.5 (Oct 01, 2025)17 benchmarksCoding 105Agents 104Math –Science 101Reasoning 10817–
139Qwen3.5 35B A3Bopen22.0 GB · 8 benchmarksCoding 86Agents –Math –Science 108Reasoning 107822.0 GB
140Qwen3 Max (Sep 23, 2025)11 benchmarksCoding –Agents –Math 105Science –Reasoning 10711–
141O1 (Dec 17, 2024)13 benchmarksCoding 104Agents –Math 104Science 104Reasoning 10213–
142Gemini 2.5 Pro Preview (May 06)4 benchmarksCoding –Agents –Math –Science 98Reasoning –4–
143MiniMax M2.1open140 GB · 4 benchmarksCoding –Agents –Math –Science –Reasoning –4140 GB
144GPT 5.4 2026 03.056 benchmarksCoding 118Agents –Math –Science –Reasoning 1016–
145Claude Sonnet 4 (May 14, 2025)16 benchmarksCoding 104Agents –Math –Science 104Reasoning 10716–
146Claude 3.7 Sonnet (Feb 19, 2025)10 benchmarksCoding –Agents –Math –Science 105Reasoning 10410–
147Qwen3 Coder 480B A35B Instructopen291 GB · 6 benchmarksCoding 101Agents 105Math –Science –Reasoning –6291 GB
148Gemini 2.5 Flash Preview (May 20)7 benchmarksCoding 100Agents –Math –Science –Reasoning 1047–
149Qwen3.5 27Bopen17.9 GB · 6 benchmarksCoding 98Agents –Math –Science –Reasoning 111617.9 GB
150DeepSeek R1 0528open412 GB · 10 benchmarksCoding 107Agents –Math –Science –Reasoning 10010412 GB
151Seed 1.64 benchmarksCoding –Agents –Math –Science –Reasoning –4–
152GLM 4.6open217 GB · 6 benchmarksCoding 96Agents 104Math –Science 100Reasoning –6217 GB
153Kimi K2 Instruct 0905open617 GB · 6 benchmarksCoding 100Agents –Math –Science –Reasoning –6617 GB
154Seed OSS 36B Instructopen24.1 GB · 3 benchmarksCoding –Agents –Math –Science –Reasoning –324.1 GB
155MiniMax M2open140 GB · 6 benchmarksCoding –Agents 105Math –Science –Reasoning –6140 GB
156Mistral Medium 3.5 128Bopen82.8 GB · 9 benchmarksCoding 102Agents –Math –Science 101Reasoning 102982.8 GB
157O3 Mini (Jan 31, 2025)19 benchmarksCoding 103Agents –Math 106Science 106Reasoning 10019–
158Grok 3 Mini Beta6 benchmarksCoding 99Agents –Math –Science –Reasoning –6–
159GLM 4.5open218 GB · 7 benchmarksCoding 99Agents 105Math –Science –Reasoning –7218 GB
160Qwen3 30B A3B Thinking 2507open19.4 GB · 7 benchmarksCoding –Agents –Math –Science 99Reasoning 94719.4 GB
161Gemini 2.5 Flash9 benchmarksCoding –Agents 94Math –Science –Reasoning 1049–
162GPT 5.2 2025 12.116 benchmarksCoding –Agents –Math –Science –Reasoning 956–
163DeepSeek V3.1open412 GB · 7 benchmarksCoding –Agents –Math –Science –Reasoning 1067412 GB
164Kimi K2 Instructopen617 GB · 10 benchmarksCoding 102Agents 102Math –Science –Reasoning 9310617 GB
165Qwen3 235B A22Bopen143 GB · 9 benchmarksCoding 101Agents –Math –Science –Reasoning 1019143 GB
166Gemini 2.5 Flash Preview (Apr 17)5 benchmarksCoding –Agents –Math –Science –Reasoning –5–
167Qwen3.5 9Bopen6.6 GB · 8 benchmarksCoding –Agents –Math –Science 105Reasoning 9786.6 GB
168GPT 5 Nano (Aug 07, 2025)17 benchmarksCoding 102Agents –Math 108Science –Reasoning 9017–
169DeepSeek R1open412 GB · 12 benchmarksCoding 100Agents –Math –Science 99Reasoning 9512412 GB
170Grok Code Fast 13 benchmarksCoding 99Agents –Math –Science –Reasoning –3–
171Grok 3 Mini3 benchmarksCoding –Agents –Math –Science –Reasoning 973–
172Qwen3 32Bopen22.6 GB · 8 benchmarksCoding 93Agents –Math –Science 96Reasoning 92822.6 GB
173DeepSeek Chat5 benchmarksCoding –Agents –Math –Science –Reasoning 865–
174Qwen3 235B A22B Instruct 2507open143 GB · 8 benchmarksCoding 102Agents –Math –Science –Reasoning 988143 GB
175Qwen3.5 4Bopen3.4 GB · 3 benchmarksCoding –Agents –Math –Science –Reasoning –33.4 GB
176GLM 4.5 Airopen67.6 GB · 4 benchmarksCoding –Agents –Math –Science –Reasoning –467.6 GB
177Grok 3 Beta5 benchmarksCoding –Agents –Math –Science –Reasoning –5–
178Qwen3 14Bopen11.9 GB · 7 benchmarksCoding –Agents –Math –Science 94Reasoning 88711.9 GB
179GPT OSS 120Bopen70.8 GB · 19 benchmarksCoding 98Agents 82Math –Science 103Reasoning 961970.8 GB
180DeepSeek R1 Distill Qwen 32Bopen24.3 GB · 3 benchmarksCoding –Agents –Math –Science –Reasoning –324.3 GB
181Qwen3 30B A3B Instruct 2507open19.4 GB · 4 benchmarksCoding –Agents –Math –Science –Reasoning 93419.4 GB
182GPT OSS 20Bopen13.1 GB · 10 benchmarksCoding 102Agents –Math –Science 93Reasoning 911013.1 GB
183QwQ 32Bopen24.3 GB · 5 benchmarksCoding –Agents –Math –Science –Reasoning –524.3 GB
184NVIDIA Nemotron 3 Super 120B A12B BF16open74.7 GB · 4 benchmarksCoding 96Agents –Math –Science 98Reasoning –474.7 GB
185GPT 4.1 (Apr 14, 2025)16 benchmarksCoding 99Agents –Math 92Science 96Reasoning 9216–
186GPT 4.5 Preview (Feb 27, 2025)10 benchmarksCoding 96Agents –Math –Science 98Reasoning 9610–
187Mercury 2.54 benchmarksCoding 95Agents –Math –Science 99Reasoning –4–
188DeepSeek v3 0324open412 GB · 12 benchmarksCoding 100Agents –Math –Science 98Reasoning 8812412 GB
189Qwen3 8Bopen7.6 GB · 7 benchmarksCoding –Agents –Math –Science 88Reasoning 8177.6 GB
190Qwen3 30B A3Bopen19.4 GB · 6 benchmarksCoding –Agents –Math –Science –Reasoning 84619.4 GB
191Grok 34 benchmarksCoding –Agents –Math –Science –Reasoning 924–
192GPT 5.4 Nano 2026 03.176 benchmarksCoding –Agents –Math 95Science –Reasoning 946–
193GPT 5.1 2025 11.137 benchmarksCoding –Agents –Math –Science 96Reasoning 917–
194O1 Mini (Sep 12, 2024)7 benchmarksCoding 91Agents –Math –Science –Reasoning 857–
195GPT 4.1 Mini (Apr 14, 2025)15 benchmarksCoding 92Agents –Math 95Science 97Reasoning 9115–
196Gemini 2.0 Flash Thinking Exp (Jan 21)6 benchmarksCoding –Agents –Math –Science 89Reasoning –6–
197Mistral Medium 25086 benchmarksCoding 92Agents –Math –Science 91Reasoning 926–
198Gemini 2.0 Flash 0016 benchmarksCoding –Agents –Math –Science –Reasoning 876–
199O1 Preview (Sep 12, 2024)5 benchmarksCoding –Agents –Math –Science –Reasoning 995–
200GPT 5.4 Mini 2026 03.176 benchmarksCoding –Agents –Math 89Science –Reasoning 966–
201Gemini 2.0 Pro Exp (Feb 05)3 benchmarksCoding –Agents –Math –Science –Reasoning –3–
202Claude 3.5 Sonnet (Oct 22, 2024)10 benchmarksCoding 99Agents 100Math –Science 88Reasoning 9710–
203Gemini 2.5 Flash Lite Preview Thinking (Jun 17)5 benchmarksCoding 95Agents –Math –Science –Reasoning 875–
204Mistral Medium 25054 benchmarksCoding –Agents –Math –Science 91Reasoning –4–
205DeepSeek v3open412 GB · 8 benchmarksCoding 97Agents –Math –Science 90Reasoning 878412 GB
206Magistral Small 2506open17.8 GB · 4 benchmarksCoding –Agents –Math –Science –Reasoning 88417.8 GB
207Qwen Max (Jan 25, 2025)4 benchmarksCoding –Agents –Math –Science –Reasoning –4–
208Mistral Large 3 675B Instruct 2512open454 GB · 6 benchmarksCoding 93Agents –Math –Science 97Reasoning 866454 GB
209Llama 4 Maverick 17B 128E Instructopen273 GB · 15 benchmarksCoding 81Agents –Math –Science 96Reasoning 8615273 GB
210Mistral Small 3.2 24B Instruct 2506open18.1 GB · 5 benchmarksCoding –Agents –Math –Science 82Reasoning –518.1 GB
211Magistral Small 2509open18.1 GB · 5 benchmarksCoding –Agents –Math –Science 83Reasoning –518.1 GB
212Gemini 1.5 Pro 0028 benchmarksCoding –Agents –Math –Science 89Reasoning 838–
213Claude 3.5 Sonnet (Jun 20, 2024)6 benchmarksCoding –Agents –Math –Science –Reasoning 936–
214Phi 4open12.5 GB · 5 benchmarksCoding –Agents –Math –Science 89Reasoning –512.5 GB
215Grok 2 (Dec 12)5 benchmarksCoding –Agents –Math –Science –Reasoning 875–
216Llama 3.1 405B Instructopen276 GB · 8 benchmarksCoding –Agents –Math –Science 85Reasoning 868276 GB
217GPT 4o (May 13, 2024)6 benchmarksCoding –Agents –Math –Science –Reasoning 906–
218Mistral Large Instruct 2411open79.8 GB · 5 benchmarksCoding –Agents –Math –Science –Reasoning 87579.8 GB
219Qwen2.5 72B Instructopen49.3 GB · 8 benchmarksCoding –Agents –Math –Science –Reasoning 88849.3 GB
220Llama 4 Scout 17B 16E Instructopen77.0 GB · 11 benchmarksCoding –Agents –Math –Science 84Reasoning 811177.0 GB
221Qwen2.5 Coder 32B Instructopen24.3 GB · 3 benchmarksCoding –Agents 84Math –Science –Reasoning –324.3 GB
222Mistral Large Instruct 2407open86.4 GB · 6 benchmarksCoding –Agents –Math –Science –Reasoning 86686.4 GB
223GPT 4o (Aug 06, 2024)7 benchmarksCoding –Agents –Math 75Science –Reasoning 777–
224Claude 3 Opus (Feb 29, 2024)9 benchmarksCoding –Agents –Math –Science –Reasoning 869–
225Qwen2.5 32B Instructopen24.3 GB · 3 benchmarksCoding –Agents –Math –Science –Reasoning –324.3 GB
226GPT 4o (Nov 20, 2024)10 benchmarksCoding 82Agents –Math –Science 82Reasoning 9010–
227Llama 3.3 70B Instructopen51.3 GB · 12 benchmarksCoding 76Agents –Math –Science 81Reasoning 831251.3 GB
228C4ai Command A 03 2025open78.7 GB · 4 benchmarksCoding –Agents –Math –Science –Reasoning 84478.7 GB
229Gemini 1.5 Flash 0025 benchmarksCoding –Agents –Math –Science –Reasoning –5–
230GPT 4 Turbo (Apr 09, 2024)6 benchmarksCoding –Agents –Math 75Science –Reasoning 936–
231Gemma 3 27B ITopen21.7 GB · 9 benchmarksCoding 70Agents –Math –Science 81Reasoning 74921.7 GB
232Mistral Small 3.1 24B Instruct 2503open18.1 GB · 5 benchmarksCoding –Agents –Math –Science 81Reasoning –518.1 GB
233Claude 3.5 Haiku (Oct 22, 2024)8 benchmarksCoding 88Agents –Math –Science 73Reasoning –8–
234Gemini 1.5 Pro 0014 benchmarksCoding –Agents –Math –Science –Reasoning 804–
235GPT 4.1 Nano (Apr 14, 2025)13 benchmarksCoding 75Agents –Math –Science 82Reasoning 7213–
236Llama 3.1 70B Instructopen51.3 GB · 7 benchmarksCoding –Agents –Math –Science –Reasoning 83751.3 GB
237Qwen2 72B Instructopen49.3 GB · 3 benchmarksCoding –Agents –Math –Science –Reasoning –349.3 GB
238GPT 4 (Jun 13)7 benchmarksCoding –Agents –Math –Science –Reasoning 897–
239Meta Llama 3 70B Instructopen51.3 GB · 4 benchmarksCoding –Agents –Math –Science –Reasoning –451.3 GB
240Gemma 3 12B ITopen10.9 GB · 7 benchmarksCoding –Agents –Math –Science 72Reasoning 66710.9 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.

Dataset. Grab the scores of every open model as a dataset on Hugging Face, including a timestamped snapshot for each refresh.