AI BriefingxAI Unveils Grok
Key point
xAI unveiled Grok, an AI model that leverages real-time information from the ๐ platform and provides witty answers.
Details
xAI has unveiled a new AI, Grok, modeled after 'The Hitchhiker's Guide to the Galaxy.' Grok aims for witty answers and a rebellious personality, and its core differentiator is leveraging real-time world knowledge through the ๐ platform. It is also designed to answer sensitive questions that other AI systems refuse.
Grok's core engine, Grok-1, is a frontier LLM with strong reasoning and coding capabilities. Key benchmark results are as follows:
- HumanEval (coding): 63.2%
- MMLU (multidisciplinary knowledge): 73%
These results surpass models that used comparable computing resources, such as GPT-3.5 and Inflection-1.
To verify real-world math problem-solving ability, Grok-1 was evaluated on the 2023 Hungarian National High School Mathematics Final Exam and scored 59% (a C grade). This is similar to Claude-2 (55%), but falls short of GPT-4 (68%).
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.