Microsoft Releases Phi-2, a Small LLM That Outperforms Llama 2 and Mistral 7B

An anonymous reader quotes a report from : Microsoft Research, the blue sky division of the software giant, [...] announced the release of its Phi-2 small language model (SML), a text-to-text AI program that is "small enough to run on a laptop or mobile device," according to a post on X. At the same time, Phi-2 with its 2.7 billion parameters (connections between artificial neurons) boasts performance that is comparable to other, much larger models including Meta's Llama 2-7B with its 7 billion parameters and even Mistral-7B, another 7 billion parameter model. Microsoft researchers also noted in their blog post on the Phi-2 release that it outperforms Google's brand new Gemini Nano 2 model despite it having half a billion more parameters, and delivers less "toxicity" and bias in its responses than Llama 2. Microsoft also couldn't resist taking a little dig at Google's now much-criticized, staged demo video for Gemini in which it showed off how its forthcoming largest and most powerful new AI model, Gemini Ultra, was able to solve fairly complex physics problems and even correct students' mistakes on them. As it turned out, even though it is likely a fraction of the size of Gemini Ultra, Phi-2 also was able to correctly answer the question and correct the student using the same prompts. However, despite these encouraging findings, there is a big limitation with Phi-2, at least for the time being: it is licensed only for "research purposes only," not commercial usage, under a custom Microsoft Research License, which further states Phi-2 may only be used for "non-commercial, non-revenue generating, research purposes." So, businesses looking to build products atop it are out of luck.

Read more of this story at Slashdot.