- xAI has released the weights and architecture of its 314 billion parameter LLM, Grok-1. It is trained from scratch by xAI using a custom training stack on top of JAX and Rust in October 2023.
- OpenAI accidentally leaked GPT-4.5 Turbo by publishing a blog post that was indexed by Bing and DuckDuckGo before it was quickly taken down. The post was an announcement for GPT-4.5 Turbo suggesting that it will be faster, more accurate, and have a larger knowledge base than its predecessor.
- Meta has introduced two 24K GPU clusters that it is using to train it's Llama 3 LLM. The clusters in total have 49,000 H100 GPUs each of which costs about $40,000. Llama 3 is expected in July.
- Apple is in talks to build Google's Gemini artificial intelligence engine into the iPhone. A deal between the firms could help Google expand the use of its AI services to more than 2 billion active Apple devices.
[link] [comments]