What fits in 32 GB of unified memory
Data updated . Last reviewed the same day, when this catalog snapshot was written.
2 catalog machines have 32 GB unified. Models are ranked against the Mac mini M4 32GB, the tightest usable pool at this sticker size (22.4 GB). The tok/s estimate uses Mac mini M4 32GB bandwidth (120 GB/s) and is not a benchmark. 10 of these fits do not clear the 24 GB tier.
Models
| Model | Best quant | Estimate | Headroom | Est. tok/s |
|---|---|---|---|---|
| Llama 3.2 1B Instruct | FP16 | 3.0 GB | 19.4 GB | 48.5 tok/s |
| Llama 3.2 3B Instruct | FP16 | 7.7 GB | 14.7 GB | 18.7 tok/s |
| Phi-3.5 mini | FP16 | 9.2 GB | 13.2 GB | 15.7 tok/s |
| Phi-4 mini | FP16 | 9.2 GB | 13.2 GB | 15.6 tok/s |
| Gemma 3 4B | FP16 | 10.3 GB | 12.1 GB | 14.0 tok/s |
| Mistral 7B Instruct v0.3 | FP16 | 17.4 GB | 5.0 GB | 8.28 tok/s |
| Qwen2.5 7B Instruct | FP16 | 18.3 GB | 4.1 GB | 7.88 tok/s |
| Gemma 4 E4B | FP16 | 19.2 GB | 3.2 GB | 7.50 tok/s |
| DeepSeek R1 Distill Llama 8B | FP16 | 19.3 GB | 3.1 GB | 7.47 tok/s |
| Llama 3.1 8B Instruct | FP16 | 19.3 GB | 3.1 GB | 7.47 tok/s |
| Qwen3 8B | FP16 | 19.7 GB | 2.7 GB | 7.33 tok/s |
| Ministral 3 8B | FP16 | 21.4 GB | 1.0 GB | 6.73 tok/s |
| Gemma 2 9B | FP16 | 22.2 GB | 0.2 GB | 6.49 tok/s |
| DeepSeek V4 Flash | Q8 | 4.5 GB | 17.9 GB | 705 tok/s |
| gpt-oss-20b | Q8 | 12.3 GB | 10.1 GB | 57.6 tok/s |
| Gemma 3 12B | Q8 | 12.5 GB | 9.9 GB | 9.59 tok/s |
| Gemma 4 12B | Q8 | 13.2 GB | 9.2 GB | 9.47 tok/s |
| Mistral Nemo 12B | Q8 | 14.4 GB | 8.0 GB | 9.21 tok/s |
| Ministral 3 14B | Q8 | 15.7 GB | 6.7 GB | 8.36 tok/s |
| Qwen3 14B | Q8 | 17.0 GB | 5.4 GB | 7.64 tok/s |
| Phi-4 | Q8 | 17.3 GB | 5.1 GB | 7.70 tok/s |
| Qwen2.5 14B Instruct | Q8 | 17.3 GB | 5.1 GB | 7.64 tok/s |
| Mistral Small 3.2 24B | Q4 | 15.7 GB | 6.7 GB | 8.37 tok/s |
| Gemma 3 27B | Q4 | 16.5 GB | 5.9 GB | 7.25 tok/s |
| Gemma 2 27B | Q4 | 16.6 GB | 5.8 GB | 7.21 tok/s |
| Qwen3.8 27B | Q4 | 17.0 GB | 5.4 GB | 7.29 tok/s |
| Gemma 4 26B-A4B | Q4 | 17.3 GB | 5.1 GB | 48.1 tok/s |
| Qwen3.6 27B | Q4 | 17.4 GB | 5.0 GB | 7.14 tok/s |
| GLM-4.7-Flash | Q4 | 18.8 GB | 3.6 GB | 68.2 tok/s |
| Qwen3 30B-A3B | Q4 | 19.4 GB | 3.0 GB | 65.8 tok/s |
| Gemma 4 31B | Q4 | 19.7 GB | 2.7 GB | 6.55 tok/s |
| Qwen3 32B | Q4 | 21.9 GB | 0.5 GB | 6.07 tok/s |
| DeepSeek R1 Distill Qwen 32B | Q4 | 22.0 GB | 0.4 GB | 6.04 tok/s |
| Qwen2.5 32B Instruct | Q4 | 22.0 GB | 0.4 GB | 6.04 tok/s |
| Qwen3.6 35B-A3B | Q4 | 22.3 GB | 0.1 GB | 65.0 tok/s |
Machines at 32 GB
- Mac mini M4 32GB
22.4 GB usable · $999
- MacBook Air M4 32GB
22.4 GB usable · $1,399
Questions
- How many models fit in 32 GB of unified memory?
- 35 of 49 fit on the tightest config at this size, the Mac mini M4 32GB, which keeps 22.4 GB usable.
- Which catalog machines have 32 GB unified?
- Mac mini M4 32GB (70% usable), MacBook Air M4 32GB (70% usable).
- Why is the ranking based on the tightest config?
- So a model listed as fitting works on every 32 GB machine here, including ones that reserve more memory for the OS. A looser machine has more headroom than the table shows.