upvote
Just to reiterate the question: QAT approximates quantization-errors in full-size fine-tuning, so that the model itself is insensitive to (or accommodating of, I'm not sure) them when quantized to the corresponding format.

Quantization to a non-corresponding format promises none of those accommodations.

reply
QAT models are trained to quantize well? I don’t get the concern here.
reply
Please read their blog post. This isn't a run of the mill QAT post trained model (like Gemma 4 QAT). It was natively trained in NVFP4, through and through.

> Most of Inkling 2 Mini’s weights were trained and are deployed natively in NVFP4, with the last four layers in MXFP8.

https://thinkingmachines.ai/news/inkling-2-mini/

reply
[flagged]
reply