AI models that fit in 64 GB of VRAM
209 of 285 open models we track fit at 4-bit (Q4_K_M) with an 8K context on a 64 GB discrete GPU. About 60.8 GB is usable. 204 of those also fit at 8-bit. Typical 64 GB cards: high-end workstations and 64 GB Macs (unified memory differs).
GPUs and Macs with 64 GB
Cards and machines in our table with exactly 64 GB of memory. Macs use unified memory, so fewer models fit than on a 64 GB discrete GPU; open the Mac page for the default macOS limits.
Mac M4 Pro 64GB (Mac)Mac M1 Max 64GB (Mac)Mac M4 Max 64GB (Mac)
Models that fit, largest first
Memory needed at 8K context. "Tight" means the model uses 90–100% of usable VRAM. How we calculate.
| Model | Params | Q4_K_M | 4-bit fit | Q8_0 | 8-bit fit |
|---|---|---|---|---|---|
| gpt oss 120b | 117B MoE | 59.5 GB | Tight | 60.1 GB | Tight |
| gpt oss safeguard 120b | 117B MoE | 59.5 GB | Tight | 60.1 GB | Tight |
| Hunyuan A13B Instruct | 80B MoE | 47.1 GB | Fits | 81.2 GB | No |
| Kimi Dev 72B | 73B | 44.6 GB | Fits | 75.1 GB | No |
| Llama 3.3 70B Instruct | 71B | 43.3 GB | Fits | 73.0 GB | No |
| DeepSeek R1 Distill Llama 70B | 71B | 43.3 GB | Fits | 73.0 GB | No |
| Llama 3.1 70B | 71B | 42.7 GB | Fits | 73.0 GB | No |
| Kimi Linear 48B A3B Instruct | 49B MoE | 28.9 GB | Fits | 49.6 GB | Fits |
| Nous Hermes 2 Mixtral 8x7B DPO | 47B MoE | 28.0 GB | Fits | 47.7 GB | Fits |
| falcon 40b | 42B | 25.2 GB | Fits | 42.9 GB | Fits |
| flint | 40B MoE | 23.6 GB | Fits | 40.3 GB | Fits |
| Maxwell V1 Flash | 37B MoE | 22.1 GB | Fits | 37.6 GB | Fits |
| Hermes 4.3 36B | 36B | 23.0 GB | Fits | 38.5 GB | Fits |
| Seed OSS 36B Instruct | 36B | 23.0 GB | Fits | 38.5 GB | Fits |
| Qwen3.6 35B A3B | 36B MoE | 22.1 GB | Fits | 36.2 GB | Fits |
| Ornith 1.5 35B A3B | 36B MoE | 21.4 GB | Fits | 36.2 GB | Fits |
| Qwen3.5 35B A3B | 36B MoE | 21.5 GB | Fits | 35.4 GB | Fits |
| Axion 1.0 35B | 35B MoE | 20.7 GB | Fits | 35.8 GB | Fits |
| Qwen AgentWorld 35B A3B | 35B MoE | 21.6 GB | Fits | 35.4 GB | Fits |
| Yi 34B 200K | 34B | 21.6 GB | Fits | 36.4 GB | Fits |
| Yi 34B | 34B | 21.6 GB | Fits | 36.4 GB | Fits |
| Yi 34B Chat | 34B | 21.6 GB | Fits | 36.4 GB | Fits |
| Yi 1.5 34B Chat | 34B | 21.6 GB | Fits | 36.4 GB | Fits |
| Nous Hermes 2 Yi 34B | 34B | 22.1 GB | Fits | 36.4 GB | Fits |
| DeepSeek R1 Distill Qwen 32B | 33B | 21.2 GB | Fits | 35.1 GB | Fits |
| Qwen3 32B | 33B | 21.1 GB | Fits | 35.1 GB | Fits |
| GLM Z1 32B 0414 | 33B | 19.5 GB | Fits | 33.4 GB | Fits |
| GLM 4 32B 0414 | 33B | 19.5 GB | Fits | 33.4 GB | Fits |
| granite 4.0 h small | 32B MoE | 18.9 GB | Fits | 32.6 GB | Fits |
| gemma 4 31B it | 31B | 20.0 GB | Fits | 33.3 GB | Fits |
| GLM 4.7 Flash | 31B MoE | 18.2 GB | Fits | 30.8 GB | Fits |
| Qwen3 Coder 30B A3B Instruct | 31B MoE | 18.7 GB | Fits | 31.7 GB | Fits |
| North Mini Code 1.0 | 30B MoE | 19.3 GB | Fits | 31.6 GB | Fits |
| Hy Memory 30B A3B | 30B MoE | 18.5 GB | Fits | 31.1 GB | Fits |
| Muse Glimmer 30B | 30B | 17.9 GB | Fits | 28.5 GB | Fits |
| granite 4.2 30b | 29B | 19.1 GB | Fits | 31.6 GB | Fits |
| Qwen3.8 27B | 28B | 16.8 GB | Fits | 28.4 GB | Fits |
| Qwen3.6 27B | 28B | 17.0 GB | Fits | 28.0 GB | Fits |
| Qwen3.5 27B | 28B | 17.0 GB | Fits | 28.0 GB | Fits |
| gemma 3 27b it | 27B | 17.3 GB | Fits | 28.7 GB | Fits |
| Arete 27B | 27B | 17.4 GB | Fits | 28.1 GB | Fits |
| MiniMax H3 Image PE | 27B | 16.8 GB | Fits | 28.0 GB | Fits |
| Trinity Mini | 26B MoE | 15.8 GB | Fits | 26.9 GB | Fits |
| GLiDE 27B | 26B | 16.2 GB | Fits | 27.2 GB | Fits |
| diffusiongemma 26B A4B it | 26B MoE | 17.1 GB | Fits | 26.4 GB | Fits |
| gemma 4 26B A4B it | 26B MoE | 17.2 GB | Fits | 26.4 GB | Fits |
| LFM2.5 24B A2B | 24B MoE | 14.3 GB | Fits | 24.5 GB | Fits |
| ERNIE 4.5 21B A3B PT | 22B | 13.5 GB | Fits | 22.6 GB | Fits |
| ERNIE 4.5 21B A3B Thinking | 22B | 13.5 GB | Fits | 22.6 GB | Fits |
| gpt oss safeguard 20b | 22B MoE | 11.8 GB | Fits | 12.3 GB | Fits |
| gpt oss 20b | 21B MoE | 11.8 GB | Fits | 12.3 GB | Fits |
| Groot 19B A3B 261023 | 19B MoE | 12.7 GB | Fits | 19.9 GB | Fits |
| Kimi VL A3B Thinking 2506 | 16B MoE | 10.8 GB | Fits | 16.8 GB | Fits |
| Kimi VL A3B Instruct | 16B MoE | 10.8 GB | Fits | 16.8 GB | Fits |
| Kimi VL A3B Thinking | 16B MoE | 10.3 GB | Fits | 17.2 GB | Fits |
| deepseek moe 16b base | 16B MoE | 12.5 GB | Fits | 18.6 GB | Fits |
| LLaDA2.1 mini | 16B MoE | 10.3 GB | Fits | 17.1 GB | Fits |
| Moonlight 16B A3B Instruct | 16B MoE | 10.8 GB | Fits | 16.8 GB | Fits |
| DeepSeek Coder V2 Lite Instruct | 16B MoE | 10.5 GB | Fits | 16.4 GB | Fits |
| DeepSeek V2 Lite | 16B MoE | 9.8 GB | Fits | 16.4 GB | Fits |
| DeepSeek V2 Lite Chat | 16B MoE | 10.5 GB | Fits | 16.4 GB | Fits |
| Qwen2.5 Coder 14B Instruct | 15B | 10.6 GB | Fits | 16.8 GB | Fits |
| DeepSeek R1 Distill Qwen 14B | 15B | 10.6 GB | Fits | 16.8 GB | Fits |
| Qwen3 14B | 15B | 10.3 GB | Fits | 16.6 GB | Fits |
| phi 4 | 15B | 10.4 GB | Fits | 16.7 GB | Fits |
| gemma 3 12b it | 12B | 8.5 GB | Fits | 13.4 GB | Fits |
| Llama Guard 4 12B | 12B | 9.1 GB | Fits | 14.2 GB | Fits |
| Llama 3.2 11B Vision Instruct | 11B | 7.4 GB | Fits | 11.6 GB | Fits |
| Llama 3.2 11B Vision | 11B | 8.0 GB | Fits | 12.4 GB | Fits |
| GLM 4.6V Flash | 10B | 6.7 GB | Fits | 10.3 GB | Fits |
| GLM 4.1V 9B Thinking | 10B | 6.7 GB | Fits | 10.3 GB | Fits |
| Step3 VL 10B | 10B | 6.5 GB | Fits | 9.9 GB | Fits |
| SenseNova Loop 9B Preview | 9.7B | 6.4 GB | Fits | 10.0 GB | Fits |
| Qwen3.5 9B | 9.7B | 6.4 GB | Fits | 10.0 GB | Fits |
| Ornith 1.5 9B | 9.7B | 6.6 GB | Fits | 10.0 GB | Fits |
| Qwen3.5 9B Base | 9.7B | 6.4 GB | Fits | 10.0 GB | Fits |
| MiMo V2.6 Distill Qwen 9B | 9.4B | 6.6 GB | Fits | 10.0 GB | Fits |
| glm 4 9b | 9.4B | 11.5 GB | Fits | 15.0 GB | Fits |
| internlm3 8b instruct | 8.8B | 6.0 GB | Fits | 9.7 GB | Fits |
| granite 4.2 8b | 8.8B | 6.8 GB | Fits | 10.5 GB | Fits |
| granite 4.1 8b | 8.8B | 6.8 GB | Fits | 10.5 GB | Fits |
| Qwen3 VL 8B Instruct | 8.8B | 6.5 GB | Fits | 9.9 GB | Fits |
| Molmo2 8B | 8.7B | 6.5 GB | Fits | 9.9 GB | Fits |
| LFM2.5 8B A1B | 8.5B MoE | 5.5 GB | Fits | 9.1 GB | Fits |
| Qwen2.5 VL 7B Instruct | 8.3B | 5.5 GB | Fits | 8.7 GB | Fits |
| UI TARS 1.5 7B | 8.3B | 5.5 GB | Fits | 8.7 GB | Fits |
| olmOCR 2 7B 1025 | 8.3B | 5.5 GB | Fits | 8.7 GB | Fits |
| olmOCR 7B 0225 preview | 8.3B | 5.5 GB | Fits | 8.7 GB | Fits |
| UI TARS 7B DPO | 8.3B | 5.5 GB | Fits | 8.7 GB | Fits |
| Qwen3 8B | 8.2B | 6.5 GB | Fits | 9.9 GB | Fits |
| DeepSeek R1 0528 Qwen3 8B | 8.2B | 6.5 GB | Fits | 9.9 GB | Fits |
| granite 3.3 8b instruct | 8.2B | 6.3 GB | Fits | 9.8 GB | Fits |
| granite 3.1 8b instruct | 8.2B | 6.3 GB | Fits | 9.8 GB | Fits |
| granite 3.2 8b instruct | 8.2B | 6.3 GB | Fits | 9.8 GB | Fits |
| Apertus 8B Instruct 2509 | 8.1B | 6.4 GB | Fits | 9.6 GB | Fits |
| Llama 3.1 8B Instruct | 8.0B | 6.2 GB | Fits | 9.6 GB | Fits |
| Llama 3.1 8B | 8.0B | 6.2 GB | Fits | 9.6 GB | Fits |
| Hermes 3 Llama 3.1 8B | 8.0B | 6.2 GB | Fits | 9.6 GB | Fits |
| DeepSeek R1 Distill Llama 8B | 8.0B | 6.2 GB | Fits | 9.6 GB | Fits |
| Meta Llama 3 8B Instruct | 8.0B | 6.2 GB | Fits | 9.6 GB | Fits |
| Meta Llama Guard 2 8B | 8.0B | 6.2 GB | Fits | 9.6 GB | Fits |
| Meta Llama 3.1 8B Instruct | 8.0B | 6.2 GB | Fits | 9.6 GB | Fits |
| Ling 3.0 tiny | 7.9B MoE | 5.5 GB | Fits | 8.7 GB | Fits |
| MiMo 7B RL | 7.8B | 6.2 GB | Fits | 9.4 GB | Fits |
| internlm2 5 7b chat | 7.7B | 6.0 GB | Fits | 9.2 GB | Fits |
| internlm2 chat 7b | 7.7B | 6.0 GB | Fits | 9.2 GB | Fits |
| Qwen2.5 7B Instruct | 7.6B | 5.5 GB | Fits | 8.7 GB | Fits |
| Qwen2.5 Coder 7B Instruct | 7.6B | 5.5 GB | Fits | 8.7 GB | Fits |
| DeepSeek R1 Distill Qwen 7B | 7.6B | 5.5 GB | Fits | 8.7 GB | Fits |
| Falcon H1R 7B | 7.6B | 5.3 GB | Fits | 8.5 GB | Fits |
| Olmo 3 7B Instruct | 7.3B | 7.3 GB | Fits | 10.3 GB | Fits |
| Olmo 3 1025 7B | 7.3B | 7.3 GB | Fits | 10.3 GB | Fits |
| Olmo 3 7B Think | 7.3B | 7.3 GB | Fits | 10.3 GB | Fits |
| Mistral 7B v0.1 | 7.2B | 5.5 GB | Fits | 8.6 GB | Fits |
| Mistral 7B Instruct v0.2 | 7.2B | 47.7 GB | Fits | 8.6 GB | Fits |
| Mistral 7B Instruct v0.1 | 7.2B | 5.5 GB | Fits | 8.6 GB | Fits |
| falcon 7b | 7.2B | 9.6 GB | Fits | 12.1 GB | Fits |
| falcon 7b instruct | 7.2B | 9.4 GB | Fits | 11.8 GB | Fits |
| granite 4.0 h tiny | 6.9B MoE | 4.6 GB | Fits | 7.5 GB | Fits |
| OLMoE 1B 7B 0924 | 6.9B MoE | 5.4 GB | Fits | 8.4 GB | Fits |
| deepseek coder 7b instruct v1.5 | 6.9B | 8.3 GB | Fits | 11.2 GB | Fits |
| deepseek coder 6.7b instruct | 6.7B | 8.3 GB | Fits | 11.1 GB | Fits |
| deepseek coder 6.7b base | 6.7B | 8.3 GB | Fits | 11.1 GB | Fits |
| Llama 2 7b hf | 6.7B | 8.3 GB | Fits | 11.1 GB | Fits |
| Yi 6B 200K | 6.1B | 4.4 GB | Fits | 7.0 GB | Fits |
| Yi 6B | 6.1B | 4.4 GB | Fits | 7.0 GB | Fits |
| Molmo2 4B | 4.9B | 4.3 GB | Fits | 6.2 GB | Fits |
| WeVisOne VL 4B | 4.8B | 4.3 GB | Fits | 6.2 GB | Fits |
| Mage VL Cowork | 4.7B | 4.3 GB | Fits | 6.2 GB | Fits |
| Qianfan OCR | 4.7B | 4.4 GB | Fits | 6.2 GB | Fits |
| Qwen3.5 4B | 4.7B | 3.7 GB | Fits | 5.3 GB | Fits |
| S5 nano | 4.7B | 3.7 GB | Fits | 5.4 GB | Fits |
| LightOnOCR 3 4B | 4.5B | 3.6 GB | Fits | 5.3 GB | Fits |
| SingVL S1 4B | 4.5B | 3.6 GB | Fits | 5.3 GB | Fits |
| Intern Decision 4B | 4.5B | 3.6 GB | Fits | 5.3 GB | Fits |
| Qwen3 VL 4B Instruct | 4.4B | 4.1 GB | Fits | 5.8 GB | Fits |
| gemma 3 4b it | 4.3B | 3.5 GB | Fits | 5.0 GB | Fits |
| medgemma 1.5 4b it | 4.3B | 3.5 GB | Fits | 5.0 GB | Fits |
| medgemma 4b it | 4.3B | 3.5 GB | Fits | 5.0 GB | Fits |
| gemma 3 4b pt | 4.3B | 3.6 GB | Fits | 5.4 GB | Fits |
| Phi 3.5 vision instruct | 4.1B | 5.8 GB | Fits | 7.6 GB | Fits |
| Qwen3 4B | 4.0B | 4.1 GB | Fits | 5.8 GB | Fits |
| Qwen3 4B Instruct 2507 | 4.0B | 4.1 GB | Fits | 5.8 GB | Fits |
| granite vision 4.1 4b | 4.0B | 3.2 GB | Fits | 4.6 GB | Fits |
| Phi 4 mini instruct | 3.8B | 4.1 GB | Fits | 5.6 GB | Fits |
| Phi 3 mini 4k instruct | 3.8B | 5.7 GB | Fits | 7.2 GB | Fits |
| Phi 3.5 mini instruct | 3.8B | 5.7 GB | Fits | 7.2 GB | Fits |
| Phi 3 mini 128k instruct | 3.8B | 5.7 GB | Fits | 7.2 GB | Fits |
| Qwen2.5 VL 3B Instruct | 3.8B | 2.8 GB | Fits | 4.0 GB | Fits |
| granite 4.2 3b | 3.7B | 3.3 GB | Fits | 4.8 GB | Fits |
| granite 4.1 3b llmon preview | 3.7B | 3.3 GB | Fits | 4.8 GB | Fits |
| Raon 1 nano | 3.5B | 3.0 GB | Fits | 4.3 GB | Fits |
| granite 4.1 3b | 3.4B | 3.2 GB | Fits | 4.6 GB | Fits |
| granite 4.0 micro | 3.4B | 3.2 GB | Fits | 4.6 GB | Fits |
| DeepSeek OCR 2 | 3.4B MoE | 2.9 GB | Fits | 4.0 GB | Fits |
| deepseek vl2 tiny | 3.4B MoE | 3.0 GB | Fits | 4.5 GB | Fits |
| Unlimited OCR 2 | 3.3B MoE | 3.0 GB | Fits | 4.0 GB | Fits |
| Unlimited OCR | 3.3B MoE | 2.9 GB | Fits | 4.0 GB | Fits |
| DeepSeek OCR | 3.3B MoE | 2.9 GB | Fits | 4.0 GB | Fits |
| granite 3.1 3b a800m instruct | 3.3B MoE | 2.9 GB | Fits | 4.3 GB | Fits |
| Llama 3.2 3B Instruct | 3.2B | 3.4 GB | Fits | 4.7 GB | Fits |
| Llama 3.2 3B | 3.2B | 3.6 GB | Fits | 5.1 GB | Fits |
| Hermes 3 Llama 3.2 3B | 3.2B | 3.4 GB | Fits | 4.7 GB | Fits |
| d1 3B | 3.1B | 2.3 GB | Fits | 3.4 GB | Fits |
| LFM2.5 VL 3B | 3.1B | 2.3 GB | Fits | 3.4 GB | Fits |
| Qwen2.5 3B Instruct | 3.1B | 2.9 GB | Fits | 4.3 GB | Fits |
| SmolLM3 3B | 3.1B | 3.0 GB | Fits | 4.3 GB | Fits |
| SmolLM3 3B Base | 3.1B | 3.0 GB | Fits | 4.3 GB | Fits |
| phi 2 | 2.8B | 4.7 GB | Fits | 5.8 GB | Fits |
| LFM2.5 2.6B | 2.7B | 2.3 GB | Fits | 3.4 GB | Fits |
| granite 3.3 2b instruct | 2.5B | 2.6 GB | Fits | 3.6 GB | Fits |
| MiniCPM5 2B | 2.5B | 2.5 GB | Fits | 3.5 GB | Fits |
| WeVisOne VL 2B | 2.4B | 2.8 GB | Fits | 3.6 GB | Fits |
| Qwen3.5 2B | 2.3B | 2.2 GB | Fits | 2.8 GB | Fits |
| SmolVLM Instruct | 2.2B | 3.0 GB | Fits | 3.8 GB | Fits |
| Qwen3 1.7B | 2.0B | 2.6 GB | Fits | 3.3 GB | Fits |
| Qwen3.8 27B DSpine b16 | 2.0B | 2.2 GB | Fits | 3.1 GB | Fits |
| d1 2B | 1.9B | 1.8 GB | Fits | 2.5 GB | Fits |
| LFM3 VL 2B | 1.9B | 1.8 GB | Fits | 2.5 GB | Fits |
| DeepSeek R1 Distill Qwen 1.5B | 1.8B | 1.9 GB | Fits | 2.7 GB | Fits |
| LFM3 1.7B | 1.7B | 1.8 GB | Fits | 2.5 GB | Fits |
| LFM3 1.7B Base | 1.7B | 1.7 GB | Fits | 2.5 GB | Fits |
| SmolLM2 1.7B Instruct | 1.7B | 3.0 GB | Fits | 3.7 GB | Fits |
| SmolLM2 1.7B | 1.7B | 3.0 GB | Fits | 3.7 GB | Fits |
| SmolLM 1.7B | 1.7B | 3.0 GB | Fits | 3.7 GB | Fits |
| granite 4.0 1b | 1.6B | 2.2 GB | Fits | 2.8 GB | Fits |
| Qwen2.5 1.5B Instruct | 1.5B | 1.9 GB | Fits | 2.7 GB | Fits |
| OLMo 2 0425 1B | 1.5B | 2.5 GB | Fits | 3.1 GB | Fits |
| deepseek coder 1.3b instruct | 1.3B | 2.8 GB | Fits | 3.3 GB | Fits |
| Llama 3.2 1B | 1.2B | 1.8 GB | Fits | 2.4 GB | Fits |
| Llama 3.2 1B Instruct | 1.2B | 1.6 GB | Fits | 2.1 GB | Fits |
| LFM2.5 1.2B Instruct | 1.2B | 1.3 GB | Fits | 1.8 GB | Fits |
| HunyuanOCR | 1.1B | 2.0 GB | Fits | 2.5 GB | Fits |
| PaddleOCR VL 1.6 | 959M | 1.3 GB | Fits | 1.7 GB | Fits |
| Qwen3.5 0.8B | 873M | 1.5 GB | Fits | 1.7 GB | Fits |
| LightOnOCR 3 0.8B | 853M | 1.5 GB | Fits | 1.7 GB | Fits |
| FastVLM 0.5B | 759M | 1.2 GB | Fits | 1.5 GB | Fits |
| Qwen3 0.6B | 752M | 1.9 GB | Fits | 2.2 GB | Fits |
| GOT OCR2 0 | 716M | 1.8 GB | Fits | 2.1 GB | Fits |
| Nemotron 3.5 Parse | 631M | 1.1 GB | Fits | 1.5 GB | Fits |
| GOT OCR 2.0 hf | 561M | 1.8 GB | Fits | 2.0 GB | Fits |
| SmolVLM2 500M Video Instruct | 507M | 1.1 GB | Fits | 1.2 GB | Fits |
| SmolVLM 500M Instruct | 507M | 1.1 GB | Fits | 1.2 GB | Fits |
| Qwen2.5 0.5B Instruct | 494M | 1.2 GB | Fits | 1.4 GB | Fits |
| LFM2.5 VL 450M | 449M | 0.8 GB | Fits | 1.0 GB | Fits |
| SmolLM2 360M Instruct | 362M | 1.1 GB | Fits | 1.2 GB | Fits |
| SmolLM2 360M | 362M | 1.1 GB | Fits | 1.2 GB | Fits |
| LFM2.5 350M | 354M | 0.8 GB | Fits | 1.0 GB | Fits |
| LFM2 350M | 354M | 1.0 GB | Fits | 1.1 GB | Fits |
Questions
How many AI models fit in 64 GB of VRAM?
209 of the 285 open models we track fit at 4-bit (Q4_K_M) with an 8K context. 204 also fit at 8-bit (Q8_0) at the same context.
What is the largest model that fits in 64 GB?
Of the models we track, the largest that fits at 4-bit with 8K context is gpt oss 120b (117B), needing about 59.5 GB.
How much of 64 GB can a model actually use?
About 60.8 GB. We treat 95% of a discrete GPU's VRAM as usable after driver and display reserve, and a model "fits" when it needs at most 90% of that. The formulas are on the methodology page.
Other VRAM sizes
A generic 64 GB discrete GPU is modelled with 95% of VRAM usable, the same rule as the rest of the site. For a specific card, including its bandwidth and estimated tokens/sec, open that GPU page. Data: Hugging Face Hub API and vendor spec pages. Report an error.