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.
| # | Model | llmrun Composite Score | Coding | Agents | Math | Science | Reasoning | Benchmarks | VRAM |
|---|---|---|---|---|---|---|---|---|---|
| 241 | Mistral Large 24023 benchmarks | Coding – | Agents – | Math – | Science – | Reasoning – | 3 | – | |
| 242 | Gemini 1.5 Flash 0014 benchmarks | Coding – | Agents – | Math – | Science – | Reasoning 74 | 4 | – | |
| 243 | Gemma 2 27B ITopen19.5 GB · 3 benchmarks | Coding – | Agents – | Math – | Science – | Reasoning – | 3 | 19.5 GB | |
| 244 | Claude 3 Sonnet (Feb 29, 2024)4 benchmarks | Coding – | Agents – | Math – | Science – | Reasoning – | 4 | – | |
| 245 | Mixtral 8x22B Instruct v0.1open88.4 GB · 5 benchmarks | Coding – | Agents – | Math – | Science – | Reasoning 78 | 5 | 88.4 GB | |
| 246 | GPT 4o Mini (Jul 18, 2024)13 benchmarks | Coding 68 | Agents – | Math 76 | Science – | Reasoning 73 | 13 | – | |
| 247 | Gemma 2 9B ITopen7.3 GB · 3 benchmarks | Coding – | Agents – | Math – | Science – | Reasoning – | 3 | 7.3 GB | |
| 248 | Claude 2.15 benchmarks | Coding – | Agents – | Math – | Science – | Reasoning 73 | 5 | – | |
| 249 | Claude 3 Haiku (Mar 07, 2024)6 benchmarks | Coding – | Agents – | Math – | Science – | Reasoning 67 | 6 | – | |
| 250 | Mixtral 8x7B Instruct v0.1open30.5 GB · 4 benchmarks | Coding – | Agents – | Math – | Science – | Reasoning 71 | 4 | 30.5 GB | |
| 251 | Mistral Nemo Instruct 2407open11.0 GB · 4 benchmarks | Coding – | Agents – | Math – | Science – | Reasoning – | 4 | 11.0 GB | |
| 252 | Qwen2.5 7B Instructopen5.8 GB · 5 benchmarks | Coding – | Agents – | Math – | Science – | Reasoning 65 | 5 | 5.8 GB | |
| 253 | Gemma 3 4B ITopen5.0 GB · 5 benchmarks | Coding – | Agents – | Math – | Science – | Reasoning 69 | 5 | 5.0 GB | |
| 254 | Llama 3.1 8B Instructopen7.9 GB · 9 benchmarks | Coding 56 | Agents – | Math – | Science 56 | Reasoning 70 | 9 | 7.9 GB | |
| 255 | Meta Llama 3 8B Instructopen7.9 GB · 5 benchmarks | Coding – | Agents – | Math – | Science – | Reasoning 65 | 5 | 7.9 GB | |
| 256 | GPT 3.5 Turbo (Jan 25)8 benchmarks | Coding – | Agents – | Math 72 | Science – | Reasoning 36 | 8 | – | |
| 257 | Llama 2 70B Chat HFopen50.2 GB · 4 benchmarks | Coding – | Agents – | Math – | Science – | Reasoning 42 | 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.
Dataset. Grab the scores of every open model as a dataset on Hugging Face, including a timestamped snapshot for each refresh.