Back

SamCtrlPlusAltMan
•

OpenAI • 1y

Alibaba has unveiled QwQ-32B-Preview, a new “open” AI model that stands out for its logical reasoning and problem-solving capabilities. With 32.5 billion parameters and the ability to process up to 32,000 words of context, QwQ is already making waves by matching—and sometimes surpassing—OpenAI’s latest models in specific benchmarks, delivering strong results in mathematical tests like AIME and MATH.

1 Reply

Replies (1)

SamCtrlPlusAltMan
•

OpenAI • 1y

Continued: It only performs well at the benchmark questions, I suppose due to data contamination + it sometimes mixes up its response with Chinese so you can't really understand the reasoning till you know the language. And the context length is far too small compared to o1 which has around 128k with an output of 66k. It's a big step forward for Open source but it still has a long way to go.

Recommendations from Medial

Account Deleted

Hey I am on Medial • 8m

Xiaomi has introduced MiMo, its first open-source large language model, developed by the newly formed Big Model Core Team. With 7 billion parameters, MiMo excels in mathematical reasoning and code generation, matching the performance of significantly

Aroneo

| Technologist | ML ... • 10m

In the ever-evolving AI landscape, a new player is making waves — Deepseek. While OpenAI, Google DeepMind, and Meta AI have been dominant forces, Deepseek is emerging as a formidable contender in the AI race.The recent buzz around Deepseek stems from