Elon Musk’s AI startup, X.ai, has unveiled its latest generative AI model, Grok-1.5. Expected to power the social network X’s Grok chatbot in the near future, Grok-1.5 shows significant improvements over its predecessor, Grok-1 based on benchmark results and specifications released by X.ai.
Grok-1.5 boasts enhanced reasoning abilities, especially in coding and math-related tasks. It has more than doubled Grok-1’s score on the popular mathematics benchmark, MATH, and has performed over ten percentage points better on the HumanEval test, which evaluates programming language generation and problem-solving skills.
One notable upgrade in Grok-1.5 is its ability to handle larger amounts of context compared to Grok-1. With a 128,000-token context, the model can process information from longer documents, enabling it to handle more complex prompts while still maintaining its instruction-following capability.
Historically, X.ai’s Grok models have stood out for their ability to address questions on sensitive topics and provide responses with a rebellious tone or even rude language upon request. It remains to be seen if Grok-1.5 introduces any changes in these areas, as X.ai has not provided any hints in the blog post.
Grok-1.5 will soon be accessible to early testers on X with added new features. Musk has hinted at features like summarizing threads and replies, and suggesting content for posts, which may be incorporated into the model soon.
The announcement of Grok-1.5 follows X.ai’s decision to open-source Grok-1 while withholding the code necessary for further training. Musk has also mentioned that more users on X’s Premium plan will have access to Grok, the chatbot, a service previously limited to Premium+ customers.
Source link
