Meta's large language model and AI assistant are getting upgrades.
On Thursday, the company released the first models of Llama 3 in two sizes, 8B and 70B parameters. They've also been integrated into Meta AI, the company's AI assistant.
"With this new model, we believe Meta AI is now the most intelligent AI assistant that you can freely use," CEO Mark Zuckerberg said Thursday in an Instagram post.
Meta said in a blog post Thursday its newest models saw "substantially reduced false refusal rates, improved alignment, and increased diversity in model responses," as well as progress in reasoning, generating code, and instruction.
"With Llama 3, we set out to build the best open models that are on par with the best proprietary models available today," the post reads. "This next generation of Llama demonstrates state-of-the-art performance on a wide range of industry benchmarks and offers new capabilities, including improved reasoning. We believe these are the best open source models of their class, period."
Though Meta bills Llama as open-source, Llama 2 required companies with more than 700 million monthly active users to request a license from the company to use it, which Meta may or may not grant.
In the near future, Meta hopes to "make Llama 3 multilingual and multimodal, have longer context, and continue to improve overall performance across core LLM capabilities such as reasoning and coding," the company said in the blog post.
So what do the changes mean for Meta AI now?
The AI assistant can help with tasks like recommending restaurants, planning trips, and making your emails sound more professional.
Using Meta AI's Imagine feature also now produces sharper images faster: They'll start to appear as you're typing and change "with every few letters typed," according to a press release issued Thursday.
Meta AI is available across Facebook, Instagram, WhatsApp, Messenger, and on browsers. The company says multimodal Meta AI is also coming to its Ray-Ban Meta smart glasses soon. It's rolling out in English in over a dozen countries outside the US.
As for what comes next, Meta says it's working on models over 400B parameters that are still in training.
"I don't think that today many people really think about Meta AI when they think about the main AI assistants that people use," Meta CEO Mark Zuckerberg told The Verge in an article published Thursday. "But I think that this is the moment where we're really going to start introducing it to a lot of people, and I expect it to be quite a major product."
Meta AI, of course, faces stiff competition from better-known AI assistants, including the likes of OpenAI's ChatGPT, Microsoft's Copilot, and Anthropic's Claude.