Can I Run Gemma 4 31B on a Apple M3 Ultra (96GB)?
Runs at full precision (fp16). Zero quality loss.
11 quantizations fit your 96.0GB
| Quant | Min VRAM | Recommended | File size | Headroom |
|---|---|---|---|---|
| fp16BEST | 63.6 GB | 65.1 GB | 0.9 GB | +32.4 GB |
| Q8_0 | 34.3 GB | 35.8 GB | 0.5 GB | +61.7 GB |
| Q6_K | 26.8 GB | 28.3 GB | 25.2 GB | +69.2 GB |
| Q5_K_M | 23.2 GB | 24.7 GB | 21.7 GB | +72.8 GB |
| Q5_K_S | 22.6 GB | 24.1 GB | 21.1 GB | +73.4 GB |
| Q4_1 | 20.6 GB | 22.1 GB | 19.1 GB | +75.4 GB |
| Q4_K_M | 20.0 GB | 21.5 GB | 18.3 GB | +76.0 GB |
| Q4_K_S | 18.9 GB | 20.4 GB | 17.4 GB | +77.1 GB |
| Q4_0 | 18.6 GB | 20.1 GB | 17.3 GB | +77.4 GB |
| Q3_K_M | 14.1 GB | 15.6 GB | 14.7 GB | +81.9 GB |
| Q3_K_S | 13.1 GB | 14.6 GB | 13.2 GB | +83.0 GB |
Try it in the cloud first
Don't want to download Gemma 4 31B just to try it? Use a hosted API or rent a GPU by the second.
Affiliate links — we earn a commission at no cost to you.
All quant variants, benchmark scores, and use-case tags.
Top-ranked open-source models that fit in 96.0GB.
FAQ
Can the Apple M3 Ultra (96GB) run Gemma 4 31B?
Yes. The Apple M3 Ultra (96GB)'s 96.0GB of unified memory is enough to run Gemma 4 31B at fp16 quantization (63.6GB required).
What's the best quantization to use?
fp16 is the highest-precision quantization that fits in your 96.0GB. It uses about 63.6GB of memory and 65.1GB recommended for comfortable inference.
What if I need more headroom for context length?
KV cache memory grows with context length. The numbers above assume a baseline 2K-4K context. For long-context use (32K+), add another 2-6GB depending on the model architecture.