Open-weight models
42 models. See local apps for which tool runs what. Parameter counts, context limits and attention layouts are read from each Hugging Face repository (updated 2026-10-17).
| Model | Org | Params | Type | Max context | Q4_K_M, 8K | Q8_0, 32K | Runs in |
|---|---|---|---|---|---|---|---|
| Gemma 4 E2B | 5.1B | dense | 128K | 4.3 GB | 6.9 GB | ||
| Ling-3.0-tiny 7.9B-A1.3B | inclusionAI (Ant Group) | 7.9B | MoE, 1.3B active | 128K | 6.0 GB | 9.9 GB | llama.cpp |
| Gemma 4 E4B | 8.0B | dense | 128K | 6.2 GB | 10.3 GB | ||
| Llama 3.1 8B | Meta | 8.0B | dense | 128K | 7.1 GB | 14.1 GB | |
| Qwen3 8B | Qwen | 8.2B | dense | 40K | 7.4 GB | 14.8 GB | Ollamallama.cppMLX (mlx-lm) |
| Qwen3.5 9B | Qwen | 9.7B | dense | 256K | 7.4 GB | 12.6 GB | |
| Gemma 4 12B | 12B | dense | 256K | 9.4 GB | 16.6 GB | ||
| Phi-4 14B | Microsoft | 15B | dense | 16K | 11.9 GB | 23.8 GB | |
| gpt-oss-20b | OpenAI | 21B | MoE, 3.6B active | 128K | 14.4 GB | 24.7 GB | |
| LFM2.5 24B-A2B | Liquid AI | 24B | MoE, 2B active | 128K | 16.3 GB | 27.9 GB | Ollamallama.cppMLX (mlx-lm) |
| Devstral Small 2 24B | Mistral AI | 24B | dense | 384K | 17.5 GB | 32.6 GB | |
| Gemma 4 26B-A4B | 26B | MoE, 4B active | 256K | 17.8 GB | 30.8 GB | ||
| Qwen3.8 27B | Qwen | 28B | dense | 256K | 19.0 GB | 33.6 GB | |
| Qwen3.6 27B | Qwen | 28B | dense | 256K | 19.0 GB | 33.6 GB | |
| Qwen3-Coder 30B-A3B | Qwen | 31B | MoE, 3.3B active | 256K | 21.0 GB | 37.6 GB | |
| GLM-4.7-Flash 30B-A3B | Z.ai | 31B | MoE, 3B active | 198K | 21.1 GB | 36.9 GB | |
| Gemma 4 31B | 31B | dense | 256K | 22.9 GB | 41.4 GB | ||
| Nemotron 3 Nano 30B-A3B | NVIDIA | 32B | MoE, 3B active | 256K | 20.9 GB | 35.8 GB | |
| Qwen3 32B | Qwen | 33B | dense | 40K | 23.8 GB | 45.4 GB | |
| Qwen3.6 35B-A3B | Qwen | 36B | MoE, 3B active | 256K | 23.8 GB | 41.0 GB | |
| Llama 3.3 70B | Meta | 71B | dense | 128K | 48.1 GB | 89.0 GB | |
| Llama 4 Scout 17B-16E | Meta | 109B | MoE, 17B active | 10240K | 71.1 GB | 126 GB | |
| GLM-4.5-Air | Z.ai | 110B | MoE, 12B active | 128K | 72.2 GB | 128 GB | LM Studiollama.cppMLX (mlx-lm) |
| gpt-oss-120b | OpenAI | 117B | MoE, 5.1B active | 128K | 74.9 GB | 130 GB | |
| Mistral Small 4 119B | Mistral AI | 119B | MoE, 6.5B active | 1024K | 76.4 GB | 132 GB | LM Studiollama.cppMLX (mlx-lm) |
| Nemotron 3 Super 120B-A12B | NVIDIA | 124B | MoE, 12B active | 256K | 79.0 GB | 137 GB | |
| Qwen3.5 122B-A10B | Qwen | 125B | MoE, 10B active | 256K | 80.0 GB | 139 GB | |
| Ling-3.0-flash 124B-A5.1B | inclusionAI (Ant Group) | 127B | MoE, 5.1B active | 256K | 81.4 GB | 141 GB | llama.cpp |
| Mistral Medium 3.5 128B | Mistral AI | 128B | dense | 256K | 84.4 GB | 153 GB | Ollamallama.cppMLX (mlx-lm) |
| Qwen3.8-Flash-Next 125B-A6B | Qwen | 180B | MoE, 6B active | 256K | 115 GB | 199 GB | |
| Step 3.5 Flash 196B-A11B | StepFun | 199B | dense | 256K | 127 GB | 221 GB | llama.cppMLX (mlx-lm) |
| MiniMax M2.7 | MiniMax | 229B | MoE | 200K | 147 GB | 260 GB | LM Studiollama.cppMLX (mlx-lm) |
| DeepSeek V4 Flash (0731) | DeepSeek | 304B | MoE, 13B active | 1024K | 193 GB | 334 GB | colibriLM Studiollama.cpp |
| MiMo-V2.6-Flash 309B-A15B | Xiaomi | 311B | MoE, 15B active | 1024K | 197 GB | 342 GB | colibrillama.cpp |
| GLM-5.3-Flash 320B-A18B | Z.ai | 321B | MoE, 18B active | 1024K | 204 GB | 353 GB | colibrillama.cppMLX (mlx-lm) |
| Qwen3-Coder 480B-A35B | Qwen | 480B | MoE, 35B active | 256K | 306 GB | 535 GB | |
| DeepSeek V3.1 | DeepSeek | 685B | MoE, 37B active | 160K | 433 GB | 753 GB | Ollamallama.cppMLX (mlx-lm) |
| GLM-5.3 | Z.ai | 753B | MoE | 1024K | 476 GB | 828 GB | colibrillama.cppMLX (mlx-lm) |
| DeepSeek V4.1 Flash | DeepSeek | 763B | MoE, 16B active | 1024K | 482 GB | 836 GB | colibri |
| Kimi K2 Instruct | Moonshot AI | 1.03T | MoE, 32B active | 128K | 648 GB | 1127 GB | llama.cppMLX (mlx-lm) |
| DeepSeek V4 Pro | DeepSeek | 1.60T | MoE, 49B active | 1024K | 1009 GB | 1751 GB | MLX (mlx-lm) |
| Kimi K3 | Moonshot AI | 2.78T | MoE, 104B active | 1024K | 1753 GB | 3044 GB | colibri |