Fine-zhtw is a Traditional Chinese (zh-TW) collection inspired by Hugging Faceās Fine series, built with mostly self-designed methods.
AI & ML interests
Traditional Chinese language models
Recent Activity
Instruction-tuned Gemma-3 models optimized for agentic workflows in Traditional Chinese.
A curated collection of datasets designed to evaluate and train reasoning capabilities in Traditional Chinese across various domains.
Benchmark log generated with Twinkle Eval, recording the model's outputs for each prompt.
Benchmark log generated with Twinkle Eval, recording the model's outputs for each prompt, see more in https://github.com/ai-twinkle/Eval
-
twinkle-ai/gemma-3-4b-it-eval-logs-and-scores
Viewer ā¢ Updated ā¢ 103k ā¢ 240 -
twinkle-ai/gemma-3-4B-T1-it-eval-logs-and-scores
Viewer ā¢ Updated ā¢ 103k ā¢ 229 -
twinkle-ai/Gemma-3-12b-it-eval-logs-and-scores
Viewer ā¢ Updated ā¢ 103k ā¢ 234 -
twinkle-ai/gemma-3-taide-12b-chat-eval-logs-and-scores
Viewer ā¢ Updated ā¢ 103k ā¢ 260
A collection of Formosa-1 (F1) reasoning models and datasets focused on Traditional Chinese instruction-following and logic.
-
twinkle-ai/Llama-3.2-3B-F1-Instruct
Text Generation ā¢ 4B ā¢ Updated ā¢ 13 ā¢ 27 -
twinkle-ai/Llama-3.2-3B-F1-Reasoning-Instruct
Text Generation ā¢ 4B ā¢ Updated ā¢ 34 ā¢ 48 -
twinkle-ai/Llama-3.2-3B-F1-Reasoning-Instruct-GGUF
Text Generation ā¢ 4B ā¢ Updated ā¢ 14 ā¢ 14 -
lianghsun/Llama-3.2-3B-F1-Base
Text Generation ā¢ 4B ā¢ Updated ā¢ 1
The official collection of Twinkle AI's visual brand assets. This repository contains our core design elements, including banners and custom emojis.
Fine-zhtw is a Traditional Chinese (zh-TW) collection inspired by Hugging Faceās Fine series, built with mostly self-designed methods.
Benchmark log generated with Twinkle Eval, recording the model's outputs for each prompt, see more in https://github.com/ai-twinkle/Eval
-
twinkle-ai/gemma-3-4b-it-eval-logs-and-scores
Viewer ā¢ Updated ā¢ 103k ā¢ 240 -
twinkle-ai/gemma-3-4B-T1-it-eval-logs-and-scores
Viewer ā¢ Updated ā¢ 103k ā¢ 229 -
twinkle-ai/Gemma-3-12b-it-eval-logs-and-scores
Viewer ā¢ Updated ā¢ 103k ā¢ 234 -
twinkle-ai/gemma-3-taide-12b-chat-eval-logs-and-scores
Viewer ā¢ Updated ā¢ 103k ā¢ 260
Instruction-tuned Gemma-3 models optimized for agentic workflows in Traditional Chinese.
A collection of Formosa-1 (F1) reasoning models and datasets focused on Traditional Chinese instruction-following and logic.
-
twinkle-ai/Llama-3.2-3B-F1-Instruct
Text Generation ā¢ 4B ā¢ Updated ā¢ 13 ā¢ 27 -
twinkle-ai/Llama-3.2-3B-F1-Reasoning-Instruct
Text Generation ā¢ 4B ā¢ Updated ā¢ 34 ā¢ 48 -
twinkle-ai/Llama-3.2-3B-F1-Reasoning-Instruct-GGUF
Text Generation ā¢ 4B ā¢ Updated ā¢ 14 ā¢ 14 -
lianghsun/Llama-3.2-3B-F1-Base
Text Generation ā¢ 4B ā¢ Updated ā¢ 1
A curated collection of datasets designed to evaluate and train reasoning capabilities in Traditional Chinese across various domains.
Benchmark log generated with Twinkle Eval, recording the model's outputs for each prompt.
The official collection of Twinkle AI's visual brand assets. This repository contains our core design elements, including banners and custom emojis.