Most big tech firms now boast fun-size versions of their flagship models for this purpose: OpenAI offers each GPT-4o and GPT-4o mini; Google DeepMind has Gemini Ultra and Gemini Nano; and Anthropic’s Claude 3 is available in three flavors: outsize Opus, midsize Sonnet, and tiny Haiku. Microsoft is pioneering a spread of small language models called Phi.
A growing variety of smaller corporations offer small models as well. The AI startup Author claims that its latest language model matches the performance of the biggest top-tier models on many key metrics despite in some cases having only a twentieth as many parameters (the values that get calculated during training and determine how a model behaves).
Smaller models are more efficient, making them quicker to coach and run. That’s excellent news for anyone wanting a more cost-effective on-ramp. And it may very well be good for the climate, too: Because smaller models work with a fraction of the pc oomph required by their giant cousins, they burn less energy.
These small models also travel well: They’ll run right in our pockets, while not having to send requests to the cloud. Small is the following big thing.
