modal-labs
/

Qwen3.8-2.4T-A95B-DFlash

Text Generation

Model card Files Files and versions

Qwen3.8-2.4T-A95B-DFlash

6-layer DFlash speculative-decoding draft for modal-labs/Qwen3.8-2.4T-A95B-NVFP4. Training data was regenerated from that model.

Downloads last month: 1,713

Safetensors

Model size

3B params

Tensor type

BF16

·

Model tree for modal-labs/Qwen3.8-2.4T-A95B-DFlash

Base model

Qwen/Qwen3.8-2.4T-A95B

Quantized

modal-labs/Qwen3.8-2.4T-A95B-NVFP4

Finetuned

(1)

this model