Welcome to Tema_Q

This is the Tema_Q development team, managed by ek15072809, Kawasumi and Amateur-Info-System. This organization focuses on developing and releasing projects related to Avant-Garde and practical Large Language Models (LLMs).

This organization posts the flagship models of Tema_Q. Please check member posts for draft models, derivative models, and quantized models.

• ek15072809
• Yuna
• Kawasumi on hiatus
• kawa3das on hiatus
• Nagisa
• Nagano
• Kazuma

For a comparison of the models' performance, please refer to the space below.

• Tema_Q Leaderboard

Please show our GitHub Repository.

• Tema_Q Agent supports tasks on local devices. It is a lightweight yet high-performance project.
• Tema_Q Browser is a simple tabbed browser designed to be operated by an LLM.
• TaQuants achieves +2-bit quality by evaluating tensor importance.
• Swap-MoE enables SSD-based on-demand inference in llama.cpp, allowing models exceeding 200B parameters to run on a system with a CPU and 16GB of DRAM.

We welcome any inquiries, including questions about licensing and usage of models and repositories, discussions about research content, and recommendations or feedback for arXiv submissions.

• X(Twitter)
• GitHub