qwen3-8b-streaming-bpw5
Near-lossless 5-bit compression (~1% perplexity; lossy) of Qwen/Qwen3-8B — produced by Sipsa Labs (UltraCompress).
- Reproducible, cryptographically verifiable reconstruction: a deterministic decode to the SHA-256-pinned validated artifact (not bit-identical to the original bf16 model).
- Independently perplexity-verified end-to-end.
- License: BUSL-1.1 + Additional Use Grant.
pip install ultracompress
Documentation & verification: https://sipsalabs.com · https://github.com/sipsalabs/ultracompress
Patents pending.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support