8 articles
A new synthetic dataset called Chimera shows that compact language models can rival far larger systems in complex reasoning tasks. Training on just 9,225 curated problems allowed a 4B model to approach the performance of models up to 50 times its size.
Alibaba's Qwen 3.5 Small Model Series arrives with benchmark results that punch well above its weight class. The 9B variant scored 90.0 in math and reached the mid-80s to low-90s across reasoning tests, making a strong case for compact, efficient AI deployment.
Nanbeige LLM Lab has released a compact 3-billion-parameter AI model that punches above its weight class, beating larger Qwen models in reasoning, coding, and agent benchmarks while handling context windows up to 256k tokens.
Alibaba's Tongyi Lab rolled out Qwen3-TTS, bringing VoiceDesign and VoiceClone to the table. These tools let you craft custom voices from text prompts and clone any voice in just three seconds, with support for ten languages and better accuracy than competitors.
New benchmark data reveals Qwen3-VL 8B Instruct outperforms the original Qwen3 across multiple tests while using just a quarter of the training resources, marking a significant efficiency breakthrough in the 8B model class.
Alibaba's Qwen3 model family has topped five major AI benchmarks, outperforming Claude Opus 4 and DeepSeek-V3.1 in reasoning, coding, and general intelligence tasks.
We use cookies to improve your experience on our site and to show you relevant advertising. To find our more, read our privacy policy and cookie policy
Saad Ullah
Marina Lyubimova
Eseandre Mordi
Artem Voloskovets
Alex Dudov
