AXIOMRIFT-3.8B

A 34-layer, approximately 3.82B-parameter Qwen3 model structurally pruned from Qwen3-4B-Instruct-2507.

Fine-tuned with QLoRA supervision on Python programs and typed symbolic intermediate representations (IR).

The LoRA adapter has been merged into the pruned BF16 checkpoint.

Downloads last month
281
Safetensors
Model size
4B params
Tensor type
BF16
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for dharun2049/AXIOMRIFT-3.8B

Adapter
(5,865)
this model
Adapters
1 model