Open-weight models
110 models with 527 quantized variants. Two numbers decide whether a machine can run a model well: the total parameter count sets how much memory you need, and the active parameter count sets how fast it generates. A sparse mixture of experts can be enormous and still fast.
All models (110)
| Model | Publisher | Type | Total | Active | Context | 4-bit size | Licence | Index |
|---|---|---|---|---|---|---|---|---|
| DeepSeek-V4 Pro reasoning · reasoning | DeepSeek | MoE | 1.6T | 49B | 1M | 871 GB | MIT | — |
| Kimi K2.6 agentic · reasoning | Moonshot AI | MoE | 1.0T | 32B | 256K | 559 GB | Modified MIT | — |
| Kimi K2 1T-A32B agentic | Moonshot AI | MoE | 1.0T | 32B | 128K | 559 GB | Modified MIT | — |
| MiMo-V2.6 Pro 1.02T-A42B agentic · reasoning | XiaomiMiMo | MoE | 1.0T | 42B | 1M | 556 GB | MIT | — |
| DeepSeek-V4.1 Flash agentic · reasoning | DeepSeek | MoE | 763.205B | 16B | 1M | 90 GB | MIT | — |
| GLM-5.3 coding · reasoning | Z.ai | MoE | 753.33B | 40B | 1M | 410 GB | MIT | — |
| DeepSeek-V3.2 general · reasoning | DeepSeek | MoE | 685.397B | 37B | 160K | 373 GB | MIT | — |
| DeepSeek V3 671B-A37B general | DeepSeek | MoE | 684.531B | 37B | 160K | 373 GB | MIT | — |
| DeepSeek R1 671B-A37B reasoning · reasoning | DeepSeek | MoE | 684.49B | 37B | 160K | 373 GB | MIT | — |
| MiniMax Text 01 456B-A46B general | MiniMax | MoE | 456.09B | 46B | 10M | 248 GB | MiniMax | — |
| MiniMax M3 agentic · reasoning | MiniMax | MoE | 427.04B | 23B | 1M | 233 GB | Apache 2.0 | — |
| Llama 3.1 405B general | Meta | Dense | 405.853B | — | 128K | 221 GB | Llama 3.1 Community | — |
| GLM-5.3 Flash coding · reasoning | Z.ai | MoE | 321.323B | 18B | 1M | 175 GB | MIT | — |
| MiMo-V2.5 310B-A15B general · reasoning | XiaomiMiMo | MoE | 310.775B | 15B | 1M | 169 GB | MIT | — |
| MiMo-V2.6 Flash 309B-A15B agentic · reasoning | XiaomiMiMo | MoE | 309B | 15B | 1M | 168 GB | MIT | — |
| DeepSeek-V4 Flash coding · reasoning | DeepSeek | MoE | 304.18B | 13B | 1M | 81 GB | MIT | — |
| Hunyuan Hy3 general · reasoning | Tencent | MoE | 298.786B | 21B | 256K | 163 GB | Tencent Hunyuan Community | — |
| DeepSeek Coder V2 236B-A21B coding | DeepSeek | MoE | 235.741B | 21B | 160K | 128 GB | DeepSeek | — |
| DeepSeek V2 236B-A21B general | DeepSeek | MoE | 235.741B | 21B | 160K | 128 GB | DeepSeek | — |
| DeepSeek V2.5 236B-A21B general | DeepSeek | MoE | 235.741B | 21B | 160K | 128 GB | DeepSeek | — |
| Qwen3 235B-A22B general · reasoning | Alibaba Qwen | MoE | 235.094B | 22B | 256K | 128 GB | Apache 2.0 | — |
| Step 3.7 Flash 198B-A11B agentic · reasoning | StepFun | MoE | 201.365B | 11B | 256K | 110 GB | Apache 2.0 | — |
| Qwen3.8 Flash Next agentic · reasoning | Alibaba Qwen | MoE | 180B | 6B | 256K | 98 GB | Qwen Community 1.0 | — |
| Mixtral 8x22B general | Mistral AI | MoE | 140.63B | 39B | 64K | 77 GB | Apache 2.0 | — |
| Qwen3.5 122B-A10B general · reasoning | Alibaba Qwen | MoE | 125.086B | 10B | 256K | 68 GB | Apache 2.0 | — |
| Nemotron 3 Super 120B-A12B agentic · reasoning | NVIDIA | MoE | 123.611B | 12B | 1M | 67 GB | NVIDIA Nemotron Open Model License | — |
| Mistral Large 2 123B general | Mistral AI | Dense | 122.61B | — | 128K | 67 GB | MRL | — |
| Mistral Small 4 119B-A6.5B general · reasoning | Mistral AI | MoE | 119.401B | 6.5B | 256K | 65 GB | Apache 2.0 | — |
| gpt-oss-120b general · reasoning | OpenAI | MoE | 116.829B | 5.1B | 128K | 63 GB | Apache 2.0 | — |
| Command R+ 104B general | Cohere For AI | Dense | 103.811B | — | 128K | 57 GB | CC-BY-NC-4.0 | — |
| Qwen2.5 72B general | Alibaba Qwen | Dense | 72.706B | — | 32K | 40 GB | Qwen | — |
| Qwen2 72B general | Alibaba Qwen | Dense | 72.706B | — | 32K | 40 GB | Qwen | — |
| Llama 3.3 70B general | Meta | Dense | 70.554B | — | 128K | 38 GB | Llama 3.3 Community License | — |
| DeepSeek R1 Distill Llama 70B reasoning · reasoning | DeepSeek | Dense | 70.554B | — | 128K | 38 GB | MIT | — |
| Llama 3.1 70B general | Meta | Dense | 70.554B | — | 128K | 38 GB | Llama 3.1 Community | — |
| Llama 3 70B general | Meta | Dense | 70.554B | — | 8K | 38 GB | Llama 3 Community | — |
| Llama 3.1 Nemotron 70B reasoning · reasoning | NVIDIA | Dense | 70.554B | — | 128K | 38 GB | NVIDIA Open Model | — |
| Llama 2 70B general | Meta | Dense | 68.977B | — | 4K | 38 GB | Llama 2 Community | — |
| Mixtral 8x7B general | Mistral AI | MoE | 46.703B | 13B | 32K | 25 GB | Apache 2.0 | — |
| Phi-3.5 MoE 42B-A6.6B general | Microsoft | MoE | 41.873B | 6.6B | 128K | 23 GB | MIT | — |
| Command R 35B general | Cohere For AI | Dense | 34.981B | — | 128K | 19 GB | CC-BY-NC-4.0 | — |
| Yi 1.5 34B general | 01.AI | Dense | 34.389B | — | 4K | 19 GB | Apache 2.0 | — |
| DeepSeek Coder 33B coding | DeepSeek | Dense | 33.343B | — | 16K | 18 GB | DeepSeek | — |
| DeepSeek-R1-Distill 32B reasoning · reasoning | DeepSeek | Dense | 32.764B | — | 128K | 18 GB | MIT | — |
| Qwen2.5 32B general | Alibaba Qwen | Dense | 32.764B | — | 32K | 18 GB | Apache 2.0 | — |
| Qwen2.5 Coder 32B coding | Alibaba Qwen | Dense | 32.764B | — | 32K | 18 GB | Apache 2.0 | — |
| Qwen3 32B general · reasoning | Alibaba Qwen | Dense | 32.762B | — | 40K | 18 GB | Apache 2.0 | — |
| Nemotron 3.5 Lightning 30B-A3B agentic · reasoning | NVIDIA | MoE | 31.578B | 3B | 1M | 17 GB | OpenMDW 1.1 | — |
| Gemma 4 31B general | Google DeepMind | Dense | 31.273B | — | 256K | 17 GB | Gemma Terms of Use | — |
| Qwen3-Coder 30B-A3B coding | Alibaba Qwen | MoE | 30.532B | 3.3B | 256K | 17 GB | Apache 2.0 | — |
| Qwen3 30B-A3B general · reasoning | Alibaba Qwen | MoE | 30.532B | 3.3B | 40K | 17 GB | Apache 2.0 | — |
| Granite 4.2 30B reasoning · reasoning | IBM | Dense | 29.277B | — | 128K | 16 GB | Apache 2.0 | — |
| Qwen3.8 27B general · reasoning | Alibaba Qwen | Dense | 27.781B | — | 256K | 15 GB | Apache 2.0 | — |
| Qwen3.6 27B coding | Alibaba Qwen | Dense | 27.781B | — | 256K | 15 GB | Apache 2.0 | — |
| Gemma 3 27B general | Google DeepMind | Dense | 27.432B | — | 128K | 15 GB | Gemma Terms of Use | — |
| Ternary Bonsai 2 27B reasoning · reasoning | Prism ML | Dense | 27.36B | — | 256K | 6 GB | Apache 2.0 | — |
| Gemma 2 27B general | Dense | 27.227B | — | 8K | 15 GB | Gemma | — | |
| Gemma 4 26B-A4B general | Google DeepMind | MoE | 25.806B | 3.8B | 256K | 14 GB | Gemma Terms of Use | — |
| Mistral Small 3.2 24B general | Mistral AI | Dense | 24.011B | — | 128K | 13 GB | Apache 2.0 | — |
| Devstral Small 24B coding | Mistral AI | Dense | 23.572B | — | 128K | 13 GB | Apache 2.0 | — |
| Codestral 22B coding | Mistral AI | Dense | 22.247B | — | 32K | 12 GB | MNPL | — |
| gpt-oss-20b general · reasoning | OpenAI | MoE | 20.915B | 3.6B | 128K | 11 GB | Apache 2.0 | — |
| InternLM 2.5 20B general | Shanghai AI Laboratory | Dense | 19.861B | — | 32K | 11 GB | InternLM | — |
| DeepSeek Coder V2 Lite 16B-A2.4B coding | DeepSeek | MoE | 15.706B | 2.4B | 160K | 9 GB | DeepSeek | — |
| DeepSeek R1 Distill Qwen 14B reasoning · reasoning | DeepSeek | Dense | 14.77B | — | 128K | 8 GB | MIT | — |
| Qwen2.5 14B general | Alibaba Qwen | Dense | 14.77B | — | 32K | 8 GB | Apache 2.0 | — |
| Qwen2.5 Coder 14B coding | Alibaba Qwen | Dense | 14.77B | — | 32K | 8 GB | Apache 2.0 | — |
| Qwen3 14B general | Alibaba Qwen | Dense | 14.768B | — | 40K | 8 GB | Apache 2.0 | — |
| Phi-4 14B reasoning | Microsoft | Dense | 14.66B | — | 16K | 8 GB | MIT | — |
| Phi-3 Medium 14B general | Microsoft | Dense | 13.96B | — | 128K | 8 GB | MIT | — |
| Llama 2 13B general | Meta | Dense | 13.016B | — | 4K | 7 GB | Llama 2 Community | — |
| Mistral NeMo 12B general | Mistral AI | Dense | 12.248B | — | 128K | 7 GB | Apache 2.0 | — |
| Gemma 3 12B vision | Dense | 12.187B | — | 128K | 7 GB | Gemma | — | |
| Llama 3.2 11B Vision vision | Meta | Dense | 10.6B | — | 128K | 6 GB | Llama 3.2 Community | — |
| Falcon3 10B general | Technology Innovation Institute | Dense | 10.306B | — | 32K | 6 GB | Falcon 3 | — |
| GLM-4 9B general | Z.ai | Dense | 9.4B | — | 128K | 5 GB | GLM-4 | — |
| Gemma 2 9B general | Dense | 9.242B | — | 8K | 5 GB | Gemma | — | |
| Yi 1.5 9B general | 01.AI | Dense | 8.829B | — | 4K | 5 GB | Apache 2.0 | — |
| Qwen2-VL 7B vision | Alibaba Qwen | Dense | 8.291B | — | 32K | 5 GB | Apache 2.0 | — |
| Qwen3 8B general · reasoning | Alibaba Qwen | Dense | 8.191B | — | 40K | 4 GB | Apache 2.0 | — |
| DeepSeek R1 Distill Llama 8B reasoning · reasoning | DeepSeek | Dense | 8.03B | — | 128K | 4 GB | MIT | — |
| Llama 3.1 8B general | Meta | Dense | 8.03B | — | 128K | 4 GB | Llama 3.1 Community | — |
| Llama 3 8B general | Meta | Dense | 8.03B | — | 8K | 4 GB | Llama 3 Community | — |
| InternLM 2.5 7B general | Shanghai AI Laboratory | Dense | 7.738B | — | 32K | 4 GB | InternLM | — |
| DeepSeek R1 Distill Qwen 7B reasoning · reasoning | DeepSeek | Dense | 7.616B | — | 128K | 4 GB | MIT | — |
| Qwen2.5 7B general | Alibaba Qwen | Dense | 7.616B | — | 32K | 4 GB | Apache 2.0 | — |
| Qwen2.5 Coder 7B coding | Alibaba Qwen | Dense | 7.616B | — | 32K | 4 GB | Apache 2.0 | — |
| Qwen2 7B general | Alibaba Qwen | Dense | 7.616B | — | 32K | 4 GB | Apache 2.0 | — |
| Phi-3 Small 7B general | Microsoft | Dense | 7.392B | — | 128K | 4 GB | MIT | — |
| Mistral 7B Instruct v0.3 general | Mistral AI | Dense | 7.248B | — | 32K | 4 GB | Apache 2.0 | — |
| Mistral 7B Instruct v0.1 general | Mistral AI | Dense | 7.242B | — | 32K | 4 GB | Apache 2.0 | — |
| DeepSeek Coder 6.7B coding | DeepSeek | Dense | 6.741B | — | 16K | 4 GB | DeepSeek | — |
| Llama 2 7B general | Meta | Dense | 6.738B | — | 4K | 4 GB | Llama 2 Community | — |
| Phi-4 Multimodal 5.6B vision | Microsoft | Dense | 5.574B | — | 128K | 3 GB | MIT | — |
| Gemma 3 4B vision | Dense | 4.3B | — | 128K | 2 GB | Gemma | — | |
| Qwen3 4B general | Alibaba Qwen | Dense | 4.022B | — | 40K | 2 GB | Apache 2.0 | — |
| Phi-4 Mini 3.8B general | Microsoft | Dense | 3.836B | — | 128K | 2 GB | MIT | — |
| Phi-3.5 Mini 3.8B general | Microsoft | Dense | 3.821B | — | 128K | 2 GB | MIT | — |
| Phi-3 Mini 3.8B general | Microsoft | Dense | 3.821B | — | 128K | 2 GB | MIT | — |
| Llama 3.2 3B general | Meta | Dense | 3.213B | — | 128K | 2 GB | Llama 3.2 Community | — |
| Qwen2.5 3B general | Alibaba Qwen | Dense | 3.086B | — | 32K | 2 GB | Qwen | — |
| Gemma 2 2B general | Dense | 2.614B | — | 8K | 1 GB | Gemma | — | |
| Qwen3 1.7B general | Alibaba Qwen | Dense | 2.032B | — | 40K | 1 GB | Apache 2.0 | — |
| Qwen2 1.5B general | Alibaba Qwen | Dense | 1.544B | — | 32K | 1 GB | Apache 2.0 | — |
| Qwen2.5 1.5B general | Alibaba Qwen | Dense | 1.544B | — | 32K | 1 GB | Apache 2.0 | — |
| Llama 3.2 1B general | Meta | Dense | 1.236B | — | 128K | 1 GB | Llama 3.2 Community | — |
| Gemma 3 1B general | Dense | 1B | — | 32K | 1 GB | Gemma | — | |
| Qwen3 0.6B general | Alibaba Qwen | Dense | 0.752B | — | 40K | 0 GB | Apache 2.0 | — |
| Qwen2 0.5B general | Alibaba Qwen | Dense | 0.494B | — | 32K | 0 GB | Apache 2.0 | — |
| Qwen2.5 0.5B general | Alibaba Qwen | Dense | 0.494B | — | 32K | 0 GB | Apache 2.0 | — |
“4-bit size” is the computed weight footprint of the default quantization, before the KV cache and runtime overhead — the model page for each shows the full memory requirement on real hardware. “Index” is the BenchLM aggregate open-weight score where published; it is one aggregator’s view, not a universal quality number, and we deliberately do not compute one of our own.