Evaluation
Thinking mode, pass@1 averaged over 1–32 seeds, same settings for all models.
Usage
Runs on vLLM on NVIDIA Hopper GPUs only (H100 / H200, sm90): the W4A8 kernel in vLLM requires compute capability 9.0.
vLLM:
vllm serve qda-lab/QDA-Qwen3.5-9B-W4A8
How it was made
Weights are 4-bit integers in groups of 128 and activations are 8-bit floating point (dynamic), quantized with GPTQ on a Hadamard-rotated copy of the model; the rotation is folded
into the weights, so there is no runtime transform.
The group scales were then fine-tuned by distillation. The recipe was chosen by QDA's automated search.
Reproduce
hf download qda-lab/qda-recipe-inputs --repo-type dataset --include "QDA-Qwen3.5-9B-W4A8/*" --local-dir inputsqda build inputs/QDA-Qwen3.5-9B-W4A8/qwen3.5-9b-w4a8.yaml
Needs the QDA library (the qda command). Recipe: qwen3.5-9b-w4a8.yaml; inputs
(calibration sample and distillation data): QDA-Qwen3.5-9B-W4A8/ in qda-lab/qda-recipe-inputs.
Takes about 6.4 h on 1× H100 (measured). The rebuild matches this checkpoint's KL to BF16 within 5 %; weights are not bit-identical.
License
Apache-2.0, the same as the base model.
Citation
TBD.