HF Model & Benchmark Matrix

24,078 variants · 14,984 scores · 120 benchmarks

Mistral-Medium-3.5-128B-NVFP4

nvidia/Mistral-Medium-3.5-128B-NVFP4

kind
nvfp4
quant
nvfp4 (4-bit)
family
mistral
modality
text-generation
root model
mistralai/Mistral-Medium-3.5-128B
provider
nvidia
license
other
language
—
context
—
library
Model Optimizer
vRAM (GB)
—
size on disk
—
downloads
29,853
likes
28
created
—
last modified
2026-07-01T15:47:29+00:00 (2026)
huggingface
https://huggingface.co/nvidia/Mistral-Medium-3.5-128B-NVFP4
tags
Model Optimizersafetensorsmistral3nvidiaModelOptMistralMistral-Medium-3.5quantizedNVFP4nvfp4text-generationconversationalbase_model:mistralai/Mistral-Medium-3.5-128Bbase_model:quantized:mistralai/Mistral-Medium-3.5-128Blicense:other8-bitmodeloptregion:us

Benchmark scores (6)

benchmarkcategoryscorecontextconfevidence
AIME 2025math88.75high| NVFP4 | 82.20% | 76.80% | 65.10% | 42.60% | 88.75% | 69.17% | 62.79% |
MMLU-Proknowledge82.2high| NVFP4 | 82.20% | 76.80% | 65.10% | 42.60% | 88.75% | 69.17% | 62.79% |
GPQA Diamondreasoning76.8high| NVFP4 | 82.20% | 76.80% | 65.10% | 42.60% | 88.75% | 69.17% | 62.79% |
IFBenchinstruction-following69.17high| NVFP4 | 82.20% | 76.80% | 65.10% | 42.60% | 88.75% | 69.17% | 62.79% |
MMMUmultimodal62.79high| NVFP4 | 82.20% | 76.80% | 65.10% | 42.60% | 88.75% | 69.17% | 62.79% |
SciCodecoding42.6high| NVFP4 | 82.20% | 76.80% | 65.10% | 42.60% | 88.75% | 69.17% | 62.79% |

Discussions (0)

none scraped.

Hugging Face model & benchmark data scraped & matrix-ified. 24,078 variants · 14,984 scores · 120 benchmarks.