ZQN-12B: our second open model, plus the M3-GPTQ compression paper

ZAQERIS Ai, an independent research group in Cairo, released ZQN-12B, its second open model, on Hugging Face along with a technical paper describing its compression method, M3-GPTQ. The paper details a technique combining Block-Hadamard decorrelation, Hessian-salience outlier selection, multi-stage 8-dimensional residual vector quantization, and Block-8 Cholesky GPTQ error propagation for extreme-bitrate LLM compression. The group describes the release as an early checkpoint rather than a frontier model, with its limitations documented on the model card.

ZQN-12B is the second open model from ZAQERIS Ai, an independent AI research group in Cairo, Egypt. Model: https://huggingface.co/ZAQERIS/ZQN-12B https://huggingface.co/ZAQERIS/ZQN-12B Alongside the model we published the technical paper behind its compression: "ZAQERIS M3-GPTQ: Hessian-Aware Multi-Stage Residual Vector Quantization for Extreme Bitrate LLM Compression". It combines Block-Hadamard decorrelation, Hessian-salience outlier selection, multi-stage 8-dimensional residual vector quantization, and Block-8 Cholesky GPTQ error propagation. Paper: https://huggingface.co/ZAQERIS/ZQN-12B/blob/main/docs/papers/ZAQERIS M3 GPTQ PAPER.md https://huggingface.co/ZAQERIS/ZQN-12B/blob/main/docs/papers/ZAQERIS M3 GPTQ PAPER.md This is an early checkpoint, not a frontier model, and we are honest about its limits on the model card. Feedback and contributions are welcome.