TAIDE

Key Research and Development Activities Timeline

2025 8/22

Gemma-3-TAIDE-12b-Chat

The Gemma-3-TAIDE-12b-Chat model has been released on the TAIDE official website and Hugging Face. Compared to the previous release of Llama3.1-TAIDE-LX-8B-Chat, it was trained by the Gemma-3-12b model, which enhances common office tasks and multi-turn conversational abilities, making it suitable for use cases that require task assistance after a chat conversation.

2025 2/11

Llama3.1-TAIDE-LX-8B-Chat Release

The Llama3.1-TAIDE-LX-8B-Chat model is released on the TAIDE official website and Hugging Face, compared to the previous release of Llama3-TAIDE-LX-8B-Chat-Alpha1, it enhances the ability to handle long texts and the decoding speed of traditional Chinese, with the model's context length increasing significantly from 8K to 131K.

2024 5/10

Continued Advancement of TAIDE Project

The follow-up TAIDE project was launched, continuing to collect diverse data through data platforms, increase multimodal model training, develop and demonstrate applications in education, healthcare, and law, continuously enhancing the model's value.

2024 5/03

TAIDE Achievement Presentation

After a year of development, the TAIDE project produced commercially usable TAIDE LX-7B, research-oriented TAIDE LX-13B, and commercially usable Llama 3-TAIDE-LX-8B-Chat-Alpha1 models. The presentation was held, inviting partners to showcase various TAIDE application achievements.

2024 4/29

Llama3-TAIDE-LX-8B Release

On April 19, 2024, Meta announced Llama3, and the TAIDE team completed model training and basic testing within four days. After obtaining consent from the National Science and Technology Council, the Llama3-based Llama 3-TAIDE-LX-8B-Chat-Alpha1 model was publicly released as a preview version.

2024 4/15

TAIDE-LX-7B Release

Released on the TAIDE official website and TAIDE Huggingface, the TAIDE-LX-7B model excels in assisting with office daily tasks such as automatic summarization, writing emails, articles, and translating between Chinese and English. The training text data used for the model is legally obtained through negotiations by the TAIDE team with government units and private publishers.

2024 2/02

AI for All Collaborative Promotional Campaign

In collaboration with the Taiwan Crossover Co-creation Association(TCCA), enabling teachers to upload data via a data platform and application workshops, aiming for future integration with teaching for diffusion, laying the foundation for combining educational application promotion.

2023 10/30

TAIDE Public Sector Application Workshop

Combining expert lectures and physical workshops, inviting central ministries to understand the technical features and application scenarios of generative AI, and what preparations are needed to integrate TAIDE, brainstorming application possibilities and proposal attempts.

2023 8/18

Participation in LLM Industry-Academia Technology Exchange Conference

The TAIDE team participated in the LLM Academia-Industry Technical Exchange Conference held at National Yang Ming Chiao Tung University, engaging with the public through technical booth presentations and introducing topics related to TAIDE's development and application.

2023 6/14

First Achievement Showcase

Using Meta Llama2 as the base, with added traditional Chinese materials for model training, the "TAIDE Project Phase One Achievement Press Conference" was held, explaining the project structure and direction, and demonstrating the TAIDE 7B model.

2023 4/27