\n<\/p>\n
Microsoft launched several new \u201copen\u201d AI models<\/a> on Wednesday, the most capable of which is competitive with OpenAI\u2019s o3-mini on at least one benchmark.<\/p>\n All of the new pemissively licensed models \u2014 Phi 4 mini reasoning, Phi 4 reasoning, and Phi 4 reasoning plus \u2014 are \u201creasoning\u201d models, meaning they\u2019re able to spend more time fact-checking solutions to complex problems. They expand Microsoft\u2019s Phi \u201csmall model\u201d family, which the company launched a year ago to offer a foundation for AI developers building apps at the edge.<\/p>\n Phi 4 mini reasoning\u00a0was trained on roughly 1 million synthetic math problems generated by Chinese AI startup DeepSeek\u2019s R1 reasoning model. Around 3.8 billion parameters in size, Phi 4 mini reasoning is designed for educational applications, Microsoft says, like \u201cembedded tutoring\u201d on lightweight devices. <\/p>\n Parameters roughly correspond to a model\u2019s problem-solving skills, and models with more parameters generally perform better than those with fewer parameters.<\/p>\n Phi 4 reasoning, a 14-billion-parameter model, was trained using \u201chigh-quality\u201d web data as well as \u201ccurated demonstrations\u201d from OpenAI\u2019s aforementioned o3-mini. It\u2019s best for math, science, and coding applications, according to Microsoft.<\/p>\n As for Phi 4 reasoning plus, it\u2019s Microsoft\u2019s previously-released Phi-4 model adapted into a reasoning model to achieve better accuracy on particular tasks. Microsoft claims that Phi 4 reasoning plus approaches the performance levels of R1, a model with significantly more parameters (671 billion). The company\u2019s internal benchmarking also has Phi 4 reasoning plus matching o3-mini on OmniMath, a math skills test.<\/p>\n Phi 4 mini reasoning, Phi 4 reasoning, and Phi 4 reasoning plus are available on the AI dev platform Hugging Face<\/a> accompanied by detailed technical reports.<\/p>\n Techcrunch event<\/p>\n \n\t\t\t\t\t\t\t\t\tBerkeley, CA<\/span> \t\t\t\t\t\t\t \u201cUsing distillation, reinforcement learning, and high-quality data, these [new] models balance size and performance,\u201d wrote Microsoft in a blog post<\/a>. \u201cThey are small enough for low-latency environments yet maintain strong reasoning capabilities that rival much bigger models. This blend allows even resource-limited devices to perform complex reasoning tasks efficiently.\u201d<\/p>\n<\/div>\n
\n\t\t\t\t\t\t\t\t\t\t\t\t\t|<\/span>
\n\t\t\t\t\t\t\t\t\t\t\t\t\tJune 5<\/span>\n\t\t\t\t\t\t\t<\/p>\n
\n\t\t\t\t\tBOOK NOW<\/span>
\n\t\t\t\t<\/a>\n\t\t\t\t\t<\/div>\n<\/p><\/div>\n<\/div>\n