MiniMax H3 on Mac mini M4 Pro 64 GB: VRAM, speed and fit
MiniMax H3 at 4-bit runs on the M4 Pro 64 GB, about 2.4 h per clip.
Video generation · 1360 × 768
Tight fit · on M4 Pro 64 GB · 4-bit (GGUF Q4_K_M · NVFP4)
96% used
- GPU memory
- 62.1 GB
- of 64.0 GB
- Time
- 2.4 h
- per clip
Where the memory goes
- Model
- 42.3 GB
- Working
- 3.0 GB
- Overhead
- 819 MB
- OS
- 16.0 GB
- Free
- 1.9 GB
- Memory
- 62.1 GB of 64.0 GB
- Output
- 1360 × 768 · 121 frames · 20 steps
- Time
- 2.4 h per clip
- Running cost
- $2.76/mo at home · $72/mo on a cloud A40
Validation: Not validated
| Check | Target | Result | Status |
|---|---|---|---|
| Fits in GPU memory | 48.0 GB usable | 46.1 GB needed | Pass |
| Headroom | 5% free | 1.9 GB free (4%) | Check |
| Time per clip | under 15 min | 2.4 h | Fails |
Targets: a clip in under 15 min, and 5% of memory left free. Calculated from published specs, not benchmarked.
The setup
- Model
- MiniMax H3
- Parameters
- 20.1B
- Precision
- 4-bit (GGUF Q4_K_M · NVFP4)
- Output
- 1360 × 768, 121 frames
- Steps
- 20
- Hardware
- M4 Pro 64 GB
- GPU can use
- 48.0 GB of 64.0 GB
Worth trying
- 720p output: Fits, 2.0 h a clip
- RTX 5090: Runs it, 5.6 min a clip
- M5 Ultra · 96 GB: Runs it, 13 min a clip
- Ryzen AI Max+ 395 · 96 GB: Runs it, 32 min a clip
Running cost
- At home
- $2.76/mo
- 126 W at the wall, $0.05 a clip
- Renting instead
- $72/mo
- A40 at $0.59/h, $0.08 a clip
4 h a day, power at $0.18 per kWh, cloud at RunPod on-demand prices as of Nov 22, 2026.
Where to get it
MiniMax H3 on similar machines
- Mac mini M4 24 GBToo big
- Mac mini M4 32 GBToo big
- Mac mini M4 Pro 24 GBToo big
- Mac mini M4 Pro 48 GBToo big
- M4 Max 128 GBRuns
- M5 Max 128 GBRuns
- M3 Ultra 96 GBRuns
- M3 Ultra 256 GBRuns
All machines: MiniMax H3 VRAM requirements. Everything that runs here: what can the Mac mini M4 Pro 64 GB run?
Other models that run on Mac mini M4 Pro 64 GB
Questions
Can MiniMax H3 run on Mac mini M4 Pro 64 GB?
MiniMax H3 at 4-bit runs on the M4 Pro 64 GB, about 2.4 h per clip.
Report VR-0P1XJFC, calculated Nov 22, 2026 from published specs, not benchmarked.