Microsoft is launching the next iteration of its lightweight AI model called Phi-3. introduced. The updated family includes Phi-3 Mini with 3.8 billion parameters, Phi-3 Small with 7 billion parameters and Phi-3 Medium with 14 billion parameters.
What Does Microsoft Phi-3 Artificial Intelligence Model Offer?
Phi-3 comes after the Phi-2 model, introduced in December 2023, was surpassed in terms of performance by models such as Meta’s Llama-3 family. came from. In the face of increasing competition, Microsoft Research has applied newer techniques to its curriculum learning approach.
The new 3.8 billion-parameter model improves on the previous Phi-2 model while using significantly fewer resources than larger language models. By Microsoft’s own benchmarks, the Phi-3 Mini, with only 3.8 billion parameters, beats both Meta’s 8 billion-parameter Llama and OpenAI’s 3.5 billion-parameter GPT-3. He leaves his past behind.
Microsoft’s statement is as follows: “We introduce phi-3-mini, a 3.8 billion parameter language model trained on 3.3 trillion tokens. This model’s overall performance is competitive with models like the Mixtral 8x7B and GPT-3.5 (e.g. phi), as measured by both academic benchmarks and internal testing. -3-mini achieves 69% in MMLU and 8.38 in MT-bench, despite being small enough to be placed in a phone.
It also provides some initial parameter scaling results with 7B and 14B models trained for 4.8T tokens, called phi-3-small and phi-3-medium. We don’t; both are significantly more capable than phi-3-mini (e.g. 75% respectively). and 78% in MMLU and 8.7 and 8.9 in MT-bench).â€
Due to its small size, the Phi-3 family is optimized for low-power devices compared to larger models. It is stated that the new model has advanced natural language processing ability directly on the smartphone. This makes the Phi-3 Mini well-suited for new applications requiring AI assistance anywhere.
