Lossless 5-bit transformer compression — 14 architectures independently PPL-verified end-to-end (0.6B-405B, dense + MoE + SSM). Hermes-3-405B 1.0066x, Mistral-7B 1.00548x, Mixtral-8x7B 1.00368x. SHA-256-verifiable bit-identical reconstruction. OpenAI-compatible API at api.sipsalabs.com. pip install ultracompress
-
Updated
May 18, 2026 - Python