What fits in 64 GB of unified memory
Data updated . Last reviewed the same day, when this catalog snapshot was written.
4 catalog machines have 64 GB unified. Models are ranked against the Mac mini M4 Pro 64GB, the tightest usable pool at this sticker size (44.8 GB). The tok/s estimate uses Mac mini M4 Pro 64GB bandwidth (273 GB/s) and is not a benchmark. 1 of these fits do not clear the 48 GB tier.
Models
| Model | Best quant | Estimate | Headroom | Est. tok/s |
|---|---|---|---|---|
| Llama 3.2 1B Instruct | FP16 | 3.0 GB | 41.8 GB | 110 tok/s |
| Llama 3.2 3B Instruct | FP16 | 7.7 GB | 37.1 GB | 42.5 tok/s |
| Phi-3.5 mini | FP16 | 9.2 GB | 35.6 GB | 35.7 tok/s |
| Phi-4 mini | FP16 | 9.2 GB | 35.6 GB | 35.6 tok/s |
| Gemma 3 4B | FP16 | 10.3 GB | 34.5 GB | 31.7 tok/s |
| Mistral 7B Instruct v0.3 | FP16 | 17.4 GB | 27.4 GB | 18.8 tok/s |
| Qwen2.5 7B Instruct | FP16 | 18.3 GB | 26.5 GB | 17.9 tok/s |
| Gemma 4 E4B | FP16 | 19.2 GB | 25.6 GB | 17.1 tok/s |
| DeepSeek R1 Distill Llama 8B | FP16 | 19.3 GB | 25.5 GB | 17.0 tok/s |
| Llama 3.1 8B Instruct | FP16 | 19.3 GB | 25.5 GB | 17.0 tok/s |
| Qwen3 8B | FP16 | 19.7 GB | 25.1 GB | 16.7 tok/s |
| Ministral 3 8B | FP16 | 21.4 GB | 23.4 GB | 15.3 tok/s |
| Gemma 2 9B | FP16 | 22.2 GB | 22.6 GB | 14.8 tok/s |
| Gemma 4 12B | FP16 | 28.7 GB | 16.1 GB | 11.4 tok/s |
| Gemma 3 12B | FP16 | 29.2 GB | 15.6 GB | 11.2 tok/s |
| Mistral Nemo 12B | FP16 | 29.4 GB | 15.4 GB | 11.1 tok/s |
| Ministral 3 14B | FP16 | 33.5 GB | 11.3 GB | 9.79 tok/s |
| Phi-4 | FP16 | 35.2 GB | 9.6 GB | 9.31 tok/s |
| Qwen3 14B | FP16 | 35.4 GB | 9.4 GB | 9.24 tok/s |
| Qwen2.5 14B Instruct | FP16 | 35.4 GB | 9.4 GB | 9.24 tok/s |
| DeepSeek V4 Flash | Q8 | 4.5 GB | 40.3 GB | 1605 tok/s |
| gpt-oss-20b | Q8 | 12.3 GB | 32.5 GB | 131 tok/s |
| Mistral Small 3.2 24B | Q8 | 26.4 GB | 18.4 GB | 10.9 tok/s |
| Gemma 4 26B-A4B | Q8 | 27.2 GB | 17.6 GB | 69.0 tok/s |
| Gemma 3 27B | Q8 | 28.7 GB | 16.1 GB | 9.51 tok/s |
| Qwen3.6 27B | Q8 | 29.1 GB | 15.7 GB | 9.55 tok/s |
| Qwen3.8 27B | Q8 | 29.6 GB | 15.2 GB | 9.40 tok/s |
| Gemma 2 27B | Q8 | 30.0 GB | 14.8 GB | 9.09 tok/s |
| GLM-4.7-Flash | Q8 | 32.3 GB | 12.5 GB | 89.2 tok/s |
| Qwen3 30B-A3B | Q8 | 33.3 GB | 11.5 GB | 85.5 tok/s |
| Gemma 4 31B | Q8 | 34.0 GB | 10.8 GB | 8.37 tok/s |
| Qwen3 32B | Q8 | 37.0 GB | 7.8 GB | 7.84 tok/s |
| DeepSeek R1 Distill Qwen 32B | Q8 | 37.0 GB | 7.8 GB | 7.84 tok/s |
| Qwen2.5 32B Instruct | Q8 | 37.0 GB | 7.8 GB | 7.84 tok/s |
| Qwen3.6 35B-A3B | Q8 | 37.1 GB | 7.7 GB | 88.7 tok/s |
| Mixtral 8x7B | Q4 | 27.5 GB | 17.3 GB | 37.4 tok/s |
| Llama 3.3 70B Instruct | Q4 | 42.5 GB | 2.3 GB | 6.42 tok/s |
Machines at 64 GB
- Framework Desktop 64GB
48.0 GB usable · $1,959
- Mac mini M4 Pro 64GB
44.8 GB usable · $1,999
- GMKtec EVO-X2 64GB
48.0 GB usable · $2,199
- Minisforum MS-S1 Max 64GB
48.0 GB usable · $2,599
Questions
- How many models fit in 64 GB of unified memory?
- 37 of 49 fit on the tightest config at this size, the Mac mini M4 Pro 64GB, which keeps 44.8 GB usable.
- Which catalog machines have 64 GB unified?
- Framework Desktop 64GB (75% usable), Mac mini M4 Pro 64GB (70% usable), GMKtec EVO-X2 64GB (75% usable), Minisforum MS-S1 Max 64GB (75% usable).
- Why is the ranking based on the tightest config?
- So a model listed as fitting works on every 64 GB machine here, including ones that reserve more memory for the OS. A looser machine has more headroom than the table shows.