CoolFace
Modelpublic

TaQuants/Tema_Q-X5-Thinking-TaQuants-GGUF

sourceHugging Faceupdated 3mo agoView on Hugging Face
0likes42downloads
Model Card

Tema_Q-X5-Thinking TaQuants

The Repository Technical Report

The TemaQ development team, team zenei, has developed a new importance matrix method called **TaQuants (Tensor-aware Adaptive Quantization)**. This model is a TaQuants version of [temaq-org/TemaQ-X5-Thinking](https://huggingface.co/temaq-org/Tema_Q-X5-Thinking) created with TaQuants v3.0.

For short-term generation, both the TaIQ2 and TaIQ3 models achieve higher accuracy than 3 bits. However, all models experience collapse during long-term text generation. Therefore, only 2 bits of GGUF are provided.