LongSpark-Qwen3-4B

LongSpark parallel drafter weights for Qwen/Qwen3-4B.

File	Contents
`model.safetensors`	641,778,432 drafter parameters in BF16 (1.28 GB).
`config.json`	Drafter architecture and configuration.

The matching Qwen3 target and tokenizer are loaded separately. Use this checkpoint with the LongSpark runtime.

License: MIT. File hashes and checkpoint provenance are recorded in release-manifest.json.

Safetensors

Model size

0.6B params

Tensor type

BF16

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Hehy/LongSpark-Qwen3-4B

Base model

Finetuned

Finetuned

this model