What fits in 36 GB of unified memory
Data updated . Last reviewed the same day, when this catalog snapshot was written.
4 catalog machines have 36 GB unified. Models are ranked against the Mac Studio M4 Max 36GB, the tightest usable pool at this sticker size (25.2 GB). The tok/s estimate uses Mac Studio M5 Max 36GB bandwidth (614 GB/s) and is not a benchmark. 0 of these fits do not clear the 32 GB tier.
Models
| Model | Best quant | Estimate | Headroom | Est. tok/s |
|---|---|---|---|---|
| Llama 3.2 1B Instruct | FP16 | 3.0 GB | 22.2 GB | 248 tok/s |
| Llama 3.2 3B Instruct | FP16 | 7.7 GB | 17.5 GB | 95.5 tok/s |
| Phi-3.5 mini | FP16 | 9.2 GB | 16.0 GB | 80.3 tok/s |
| Phi-4 mini | FP16 | 9.2 GB | 16.0 GB | 80.0 tok/s |
| Gemma 3 4B | FP16 | 10.3 GB | 14.9 GB | 71.4 tok/s |
| Mistral 7B Instruct v0.3 | FP16 | 17.4 GB | 7.8 GB | 42.4 tok/s |
| Qwen2.5 7B Instruct | FP16 | 18.3 GB | 6.9 GB | 40.3 tok/s |
| Gemma 4 E4B | FP16 | 19.2 GB | 6.0 GB | 38.4 tok/s |
| DeepSeek R1 Distill Llama 8B | FP16 | 19.3 GB | 5.9 GB | 38.2 tok/s |
| Llama 3.1 8B Instruct | FP16 | 19.3 GB | 5.9 GB | 38.2 tok/s |
| Qwen3 8B | FP16 | 19.7 GB | 5.5 GB | 37.5 tok/s |
| Ministral 3 8B | FP16 | 21.4 GB | 3.8 GB | 34.4 tok/s |
| Gemma 2 9B | FP16 | 22.2 GB | 3.0 GB | 33.2 tok/s |
| DeepSeek V4 Flash | Q8 | 4.5 GB | 20.7 GB | 3609 tok/s |
| gpt-oss-20b | Q8 | 12.3 GB | 12.9 GB | 295 tok/s |
| Gemma 3 12B | Q8 | 12.5 GB | 12.7 GB | 49.1 tok/s |
| Gemma 4 12B | Q8 | 13.2 GB | 12.0 GB | 48.5 tok/s |
| Mistral Nemo 12B | Q8 | 14.4 GB | 10.8 GB | 47.1 tok/s |
| Ministral 3 14B | Q8 | 15.7 GB | 9.5 GB | 42.8 tok/s |
| Qwen3 14B | Q8 | 17.0 GB | 8.2 GB | 39.1 tok/s |
| Phi-4 | Q8 | 17.3 GB | 7.9 GB | 39.4 tok/s |
| Qwen2.5 14B Instruct | Q8 | 17.3 GB | 7.9 GB | 39.1 tok/s |
| Mistral Small 3.2 24B | Q4 | 15.7 GB | 9.5 GB | 42.8 tok/s |
| Gemma 3 27B | Q4 | 16.5 GB | 8.7 GB | 37.1 tok/s |
| Gemma 2 27B | Q4 | 16.6 GB | 8.6 GB | 36.9 tok/s |
| Qwen3.8 27B | Q4 | 17.0 GB | 8.2 GB | 37.3 tok/s |
| Gemma 4 26B-A4B | Q4 | 17.3 GB | 7.9 GB | 246 tok/s |
| Qwen3.6 27B | Q4 | 17.4 GB | 7.8 GB | 36.5 tok/s |
| GLM-4.7-Flash | Q4 | 18.8 GB | 6.4 GB | 349 tok/s |
| Qwen3 30B-A3B | Q4 | 19.4 GB | 5.8 GB | 337 tok/s |
| Gemma 4 31B | Q4 | 19.7 GB | 5.5 GB | 33.5 tok/s |
| Qwen3 32B | Q4 | 21.9 GB | 3.3 GB | 31.1 tok/s |
| DeepSeek R1 Distill Qwen 32B | Q4 | 22.0 GB | 3.2 GB | 30.9 tok/s |
| Qwen2.5 32B Instruct | Q4 | 22.0 GB | 3.2 GB | 30.9 tok/s |
| Qwen3.6 35B-A3B | Q4 | 22.3 GB | 2.9 GB | 332 tok/s |
Machines at 36 GB
- Mac Studio M4 Max 36GB
25.2 GB usable · $1,999
- Mac Studio M5 Max 36GB
25.2 GB usable · $2,499
- MacBook Pro 14-inch M4 Max 36GB
25.2 GB usable · $3,199
- MacBook Pro 16-inch M4 Max 36GB
25.2 GB usable · $3,499
Questions
- How many models fit in 36 GB of unified memory?
- 35 of 49 fit on the tightest config at this size, the Mac Studio M4 Max 36GB, which keeps 25.2 GB usable.
- Which catalog machines have 36 GB unified?
- Mac Studio M4 Max 36GB (70% usable), Mac Studio M5 Max 36GB (70% usable), MacBook Pro 14-inch M4 Max 36GB (70% usable), MacBook Pro 16-inch M4 Max 36GB (70% usable).
- Why is the ranking based on the tightest config?
- So a model listed as fitting works on every 36 GB machine here, including ones that reserve more memory for the OS. A looser machine has more headroom than the table shows.