How much memory does MiniMax M3 need in Q4?
MiniMax M3 has 428 billion parameters. The Q4 format decides how much memory is needed to load them.
How is this footprint calculated?
MiniMax M3 totals 428 billion parameters. The Q4 format takes 0.6 byte per parameter (taken here as llama.cpp Q4_K_M, 0.6 byte per parameter including scales; Q4_K_S weighs 0.56, AWQ and GPTQ 0.55). The product gives the weights, to which a 15% runtime margin is added for activations and buffers: 428 billion parameters × 0.6 byte (Q4) = 256.8 GB of weights; × 1.15 runtime margin = 295.3 GB. The attention cache is not calculated on this page. These models’ architectures (compressed latent, linear or sparse attention, Mamba layers) share no per-token formula; it is measured on the machine, with the target context and concurrency [TO BE MEASURED].
What does the format change?
Q4 brings block quantisation, broadly supported.
| Format | Weights in memory | Compatible platforms |
|---|---|---|
| FP16 | 984 GB | 4 |
| FP8 | 492 GB | 7 |
| NVFP4 | 246 GB | 7 |
| Q4 | 295 GB | 7 |
Which platforms qualify?
- Mac Studio Ultra, 512 GB
- DGX Station, 748 GB
- RTX PRO 6000 server, 768 GB
- H200 SXM server, 1,128 GB
- B200 SXM, 1,440 GB
- B300 SXM, 2,304 GB
See the MiniMax M3 fact sheet.
How much memory for MiniMax M3 in Q4?
About 295 GB for the weights, excluding the attention cache.
Which format should I choose for MiniMax M3?
Q4 brings block quantisation, broadly supported. The most precise format that fits the target machine remains the best choice.
Method and sources
Memory footprints are calculated, not measured: a reading on real hardware may differ depending on the engine and the exact weight format. Every figure below carries its source and the date of its last check.
- MiniMax M3: 428 billion parameters, 23 billion active, 1,048,576-token context, MiniMax Community licence. Hugging Face repository (config.json, API, model card), checked on 2 September 2026. About 428 billion announced, 23 active; the BF16 repository holds 427 billion elements (854 GB).
- Bytes per parameter: FP16 2 (16 bits per parameter, hence 2 bytes); FP8 1 (8 bits per parameter (E4M3), hence 1 byte, block scales not counted); NVFP4 0.5 (4 bits per parameter, hence 0.5 byte; 4-bit published weights weigh 0.54 to 0.56 byte per parameter with their scales (DeepSeek V4 Pro 865 GB for 1,599 billion, Kimi K3 1,561 GB for 2,780 billion)); Q4 0.6 (taken here as llama.cpp Q4_K_M, 0.6 byte per parameter including scales; Q4_K_S weighs 0.56, AWQ and GPTQ 0.55). Source: connaissance/faits.yaml, quantifications family, checked on 2026-08-31.
- Runtime margin × 1.15: QDNA operating assumption, not measured: 15% above the weights for activations, buffers and fragmentation.
- “Fits” threshold at 75% of memory: QDNA assumption, not measured, which keeps the remaining quarter for the attention cache and concurrency.