sunzhiyuan/
suntray-qwen3:1.5b

24 3 months ago

Suntray-Qwen3 is fine-tuned from Qwen-3 using LoRA, making it the first hybrid reasoning model that integrates fast thinking (System 1) and slow thinking (System 2). Its performance surpasses leading open-source models.

1.5b

3 months ago

0593a1b43925 · 3.6GB ·

model
qwen2
·
1.78B
·
F16

Readme

“This model is built upon Qwen-3 as its foundation and is specifically optimized for core tasks such as text representation, retrieval, and ranking. It achieves a 40% performance improvement over the previous version and has surpassed top-tier models from Google, OpenAI, and Microsoft on authoritative benchmarks like MTEB, securing SOTA (State-of-the-Art) results in its category.”