Meta’s Llama 3 Proves a Hit With Developers; Snowflake Gets In The LLM Game

It’s been five days since the release of Meta Platforms’ highly-anticipated Llama 3 models, and the new large language models have seemingly won the mindshare of AI developers—so much so that barely anyone was talking about new model releases from Microsoft, Adobe and Amazon on Monday.

A number of developers told me that they’re already experimenting with the two smaller Llama 3 models Meta released and, after getting great results, are considering switching over to them. The biggest, most powerful Llama 3 model won’t be out until the summer. But the first two are plenty powerful. One founder pointed out to me that, if you filter for performance in English on LMSYS’ leaderboard, only OpenAI's GPT-4 Turbo beats the middle version of Llama 3, the 70 billion parameter version. That augurs well for the bigger one.