Microsoft has introduced Phi-2, a 2.7 billion-parameter AI model showcasing performance in language understanding, reasoning, math, and coding abilities, claiming to surpass larger models by up to 25x its size.
Phi-2, available on Microsoft Azure AI Studio's model catalog, is designed with curated "textbook-quality" data, allowing it to excel without relying on reinforcement learning or instructional fine-tuning. It is also claimed to outperform larger models in mitigating bias and toxicity.
Part of a series of "small language models" (SLMs), Phi-2 follows Phi-1 and Phi-1.5 releases, positioning itself as an efficient platform for researchers exploring AI safety, interpretability, and ethical language model development.
By using this site, you agree to allow SPEEDA Edge and our partners to use cookies for analytics and personalization. Visit our privacy policy for more information about our data collection practices.