MBZUAI and G42 Unveil K2 Think: A Compact AI Reasoning Model

The Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) and G42 have announced the launch of K2 Think, a new open-source AI reasoning model. Announced in a press release, K2 Think is designed to deliver performance comparable to leading reasoning models while being significantly smaller, with just 32 billion parameters.

K2 Think employs innovative techniques such as long chain-of-thought supervised fine-tuning and reinforcement learning with verifiable rewards to enhance its reasoning capabilities. The model is also optimized for Cerebras' wafer-scale compute platform, achieving a throughput of 2000 tokens per second, making it one of the fastest reasoning systems available.

This model is part of a broader effort by the UAE to position itself as a leader in AI technology, showcasing the potential of smaller, more efficient models in the field of advanced AI reasoning. K2 Think is available for researchers and developers on platforms like Hugging Face, promoting transparency and collaboration in AI research.