Download
4 variants available
bf16 SafeTensor
minimax_h3_turbo_fl2va_pruned_bf16.safetensors
BF16, good balance • 37.46 GB
Verified: 5 hours ago
SafeTensor
mxfp8
minimax_h3_turbo_fl2va_pruned_mxfp8.safetensors
Block-scaled 8-bit, near-FP16 quality
Verified: 3 hours ago
int8
minimax_h3_turbo_fl2va_pruned_int8_convrot.safetensors
8-bit integer, smaller file
Verified: 6 hours ago
nvfp4
minimax_h3_turbo_fl2va_pruned_nvfp4.safetensors
NVFP4 4-bit, Blackwell-optimized
Verified: 5 hours ago
1180 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9
(11)
Sep 20, 2026
MiniMax H3


70.1K0 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9.0 1 2 3 4 5 6 7 8 9K
653.3K0 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9.0 1 2 3 4 5 6 7 8 9K
5.1M0 1 2 3 4 5 6 7 8 9.0 1 2 3 4 5 6 7 8 9M

MiniMax H3 is licensed by MiniMax under the MiniMax H3 Community License Agreement. That agreement’s Applicable Territory excludes the European Union, the United Kingdom, the Republic of Korea and the United States of America. Your use of H3 and of any H3 derivative is subject to that agreement and its Acceptable Use Policy.
MiniMax H3
Base MinimaxH3 (Turbo Merge)
Unpruned is in INT8 only as I do not feel the need to quant that to the lower quality.
This model is unchanged/untrained from the base model listed. The model was quantized from the source FP32/BF16 model.
In testing I have found this model to be 10% faster even with the high quality 64 Convrot size, then using the turbo lora.
