Microsoft Phi-3: Cheaper, smaller, yet better AI?
OpenAI has been rapidly transforming the world with ChatGPT and later, GPT-4.
Consequently, people couldn’t keep their excitement for the upcoming version.
Yet, TechCrunch reported its perplexing statement at an MIT event last year. “I think we’re at the end of the era where it’s going to be these giant models, and we’ll make them better in other ways,” it said.
READ: Microsoft says China will use AI to disrupt elections
Surprisingly, AI competitor Microsoft realized that prediction with its Phi-3 family of AI models.
The official blog says they are the “most capable and cost-effective small language models (SLMs) available.”
What are Microsoft Phi-3’s features?
We're excited to announce the launch of Phi-3, a groundbreaking family of small language models that outperform larger models on a range of benchmarks. Learn how these small language models trained on high-quality data are doing more with less: https://t.co/dFfyktuEUL
— Microsoft (@Microsoft) April 23, 2024
Microsoft announced the first of the Phi-3 family members: Phi-3-mini. It has a capacity of 3.8 billion parameters, which means it performs better than models twice its size.
Phi-3 also exceeds similar models across language, reasoning, math and coding benchmarks.
It has two context-length variants: 4K and 128K tokens. The latter represents the number of words an AI model can process.
Moreover, it is the first of its class to support 128K tokens for its context window with little impact on quality.
In other words, Microsoft Phi-3 can handle roughly that number of words without reducing performance.
The latest model is instruction-tuned, meaning it follows instructions similar to how people usually communicate. As a result, the model is ready to use out of the box.
Microsoft says its latest AI program is a more practical choice for those who want to build generative AI applications.
Moreover, it is ideal for the following purposes:
- Transforming businesses with generative AI
- Resource-constrained environments, including on-device and offline inference scenarios
- Latency-bound scenarios where fast response times are critical
- Cost-constrained use cases, particularly those with simpler tasks
Soon, IT experts in third world countries like the Philippines may use Microsoft Phi-3 to build generative AI apps.
The Bill Gates-founded company stated it will launch additional models in the coming weeks: Phi-3-small (7B) and Phi-3-medium (14B).
