Can MiniMax H3 Image PE run on the RTX 4090?
Yes. At 8K context the best fit is Q5_K_M, using about 19.3 GB of the 22.8 GB usable on the RTX 4090. Estimated speed ceiling: ~52 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) • | 11.1 GB Fits | 11.3 GB Fits | 12.8 GB Fits | 18.8 GB Fits | 94 |
| 3-bit (Q3_K_M) • | 13.5 GB Fits | 13.8 GB Fits | 15.3 GB Fits | 21.3 GB Tight | 76 |
| 4-bit (Q4_K_M) • | 16.5 GB Fits | 16.8 GB Fits | 18.3 GB Fits | 24.3 GB No | 61 |
| 5-bit (Q5_K_M) • | 19.0 GB Fits | 19.3 GB Fits | 20.8 GB Tight | 26.8 GB No | 52 |
| 6-bit (Q6_K) • | 21.7 GB Tight | 21.9 GB Tight | 23.4 GB No | 29.4 GB No | 46 |
| 8-bit (Q8_0) • | 27.8 GB No | 28.0 GB No | 29.5 GB No | 35.5 GB No | — |
| 16-bit (F16/BF16) | 52.1 GB No | 52.3 GB No | 53.8 GB No | 59.8 GB No | — |
• real GGUF file size from mradermacher/MiniMax-H3-Image-PE-GGUF. Speed is a bandwidth ceiling (1008 GB/s), not a benchmark. Methodology.
More
MiniMax H3 Image PE on every GPU and Mac · Every model that runs on the RTX 4090 · Other models that fit 24 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (MiniMaxAI/MiniMax-H3-Image-PE), NVIDIA specs. Data as of 2026-11-17.