All articles tagged with #phi 2
Originally Published 2 years ago — by Slashdot
Microsoft Research has released Phi-2, a small language model (SML) with 2.7 billion parameters that outperforms larger models like Meta's Llama 2-7B and Mistral-7B. Phi-2 also surpasses Google's Gemini Nano 2 model in performance and exhibits less toxicity and bias. However, Phi-2 is currently only licensed for research purposes, limiting its commercial usage.
Originally Published 2 years ago — by The Information
French AI startup Mistral has released an open-source model called Mixtral 8x7B, which combines smaller models to efficiently handle specific tasks and match the quality of larger models like GPT-3.5. Microsoft researchers also unveiled their mobile-friendly model, Phi-2, with just 2.7 billion parameters compared to Mixtral's 7 billion. These developments highlight the growing trend of creating smaller language models that can run on less powerful devices while still generating accurate results.
Originally Published 2 years ago — by Decrypt
Microsoft Research has released Phi-2, a small language model (SLM) with 2.7 billion parameters that demonstrates impressive performance for its size. Unlike large language models (LLMs) such as GPT and Gemini, SLMs are trained on a limited dataset with fewer parameters, making them more efficient for specific tasks like math and calculations. Phi-2 outperforms larger models in benchmarks for math, coding, and common sense reasoning. Microsoft's approach to AI also includes the development of custom chips, Maia and Cobalt, which support their vision of integrating AI and cloud computing. Phi-2's small size allows it to run locally on low-tier equipment, potentially paving the way for new applications and use cases. Its availability in the Azure AI Studio model catalog contributes to the democratization of AI research.