Elon Musk unveils Grok, xAI's advanced LLM, to rival OpenAI's GPT and Anthropic's Claude 2
Elon Musk has unveiled xAI's leading AI product, "Grok," a Large Language Model (LLM) that offers real-time data, efficiency, and "humor". The name "Grok" implies understanding and puts it in direct competition with other LLMs such as OpenAI's GPT and Anthropic's Claude 2. For now, Grok is accessible to a select group of users in the United States, with a waitlist for early access through the X social network. The pricing details are yet to be announced.
Grok-0, the initial prototype, was developed using 33 billion data parameters, which is more than the 20 billion parameters used by OpenAI's GPT-3.5, but less than newer models like Meta's Meta Llama, which uses 70 billion. Despite this, xAI claims that Grok-0 equals Meta LLama 2's capabilities with just half the training resources.
The xAI team's advancements in reasoning and coding led to the creation of Grok-1, the advanced LLM that drives the Grok chatbot client. In its computing class, Grok surpasses other models like ChatGPT -3.5 and Inflection-1 in machine learning benchmarks and tasks. Future plans include making Grok available to all X Premium+ subscribers, although the timeline has not been disclosed.

Comments
Gpt 3.5 uses 100 million parameters, and Gpt-4 uses 1 trillon.
100 billion for Gpt 3.5 ** and more precisely 175 billion
To be fair, the thing with GPT-3.5 is that OpenAI hasn't provided an official statement, as far as I know. And recently a Microsoft codediffusion paper suggests it actually has 20 billion parameters, though this is still being debated:
https://www.reddit.com/r/LocalLLaMA/comments/17jrj82/new_microsoft_codediffusion_paper_suggests_gpt35/
https://www.reddit.com/r/LocalLLaMA/comments/17lvquz/clearing_up_confusion_gpt_35turbo_may_not_be_20b/
Thank you very much it’s good to know!
How is 33 billion less than 20 billion
Love that he poked fun at "Sorry, I’m a language model and can’t answer that" that current AI love to recite.
It's pretty annoying to be hit with a brick wall, even with a pretty safe questions.