Model techniques map
Taxonomyoptimizationquantization-aware training

taxonomy node · level 2

quantization-aware training

15 methods filed at this node or below it, from the sources of 10 models.

optimization :: quantization-aware training

Matching aids for the classifier: QAT; FP4 QAT; INT4 QAT; straight-through estimator; stochastic rounding.

In this branch 15

Everything filed at this node or below it, with one collapsible heading per child node.

filed here 15

Quantization-Aware Training core · 8 sources · 8 quotes
Quantize-Dequantize Training core · 1 source · 1 quote
FP4 Quantization default · 4 sources · 4 quotes
FP4 Quantization-Aware Training used · 3 sources · 3 quotes
MXFP4 Weights with MXFP8 Activations used · 2 sources · 3 quotes
Stochastic Rounding used · 2 sources · 3 quotes
INT4 Quantization-Aware Training used · 1 source · 1 quote
MXFP4 Quantization-Aware Post-Training used · 1 source · 1 quote
NVFP4 Training used · 1 source · 1 quote
Per-Block Scalar Scaling used · 1 source · 1 quote
Q4_0 Quantization Format used · 1 source · 1 quote
Random Hadamard Transforms used · 1 source · 1 quote
Stochastic Rounding for Mamba Cache used · 1 source · 1 quote
Stochastic Rounding of Gradients used · 1 source · 1 quote
Straight-Through Estimator used · 1 source · 1 quote

By model

Which of this branch's techniques each model's own documents describe, and how strongly. Under each model: its strongest treatment anywhere in the branch.

Modeltechniques
NVIDIA-Nemotron-3-Ultra-550B-A55B usedStochastic Rounding usedStochastic Rounding for Mamba Cache used—
MiMo-V2.6-Flash coreQuantize-Dequantize Training coreFP4 Quantization-Aware Training used—
DeepSeek-V4-Flash usedFP4 Quantization usedFP4 Quantization-Aware Training usedQuantization-Aware Training usedStraight-Through Estimator used—
GLM-5.2 usedINT4 Quantization-Aware Training used—
DeepSeek-V4-Pro usedFP4 Quantization usedFP4 Quantization-Aware Training usedQuantization-Aware Training usedStraight-Through Estimator used—
Gemma 4 31B coreQuantization-Aware Training corePer-Block Scalar Scaling usedQ4_0 Quantization Format used—
Kimi K3 usedMXFP4 Quantization-Aware Post-Training usedMXFP4 Weights with MXFP8 Activations usedQuantization-Aware Training used—
MiMo-V2.6-Pro coreQuantize-Dequantize Training coreFP4 Quantization-Aware Training used—
NVIDIA-Nemotron-3.5-Lightning-30B-A3B usedNVFP4 Training usedRandom Hadamard Transforms usedStochastic Rounding of Gradients used—
gpt-oss-120b defaultFP4 Quantization default—