xAI, an Artificial Intelligence (AI) company, founded by Elon Musk, has announced the release of Grok-1.5, the latest iteration of its large language model (LLM).
Grok-1.5, xAI said, boasts significant improvements in reasoning capabilities and the ability to understand long contexts.
Grok-1.5 demonstrates advanced problem-solving skills, particularly in coding and math tasks, according to the company which said achieved high scores on several benchmarks, surpassing other models in its compute class.
- MATH benchmark: 50.6% accuracy
- GSM8K benchmark: 90% accuracy (evaluates middle and high school math problems)
- HumanEval benchmark: 74.1% accuracy (evaluates code generation and problem-solving)
According to xAI, in terms of long context understanding, Grok-1.5 can now process information within a context window of up to 128,000 tokens, which is 16 times more than the previous version demonstrating an increment in memory capabilities. This allows it to handle complex prompts and utilize information from lengthy documents.
xAI revealed that Grok-1.5 will be available soon to early testers on the 𝕏 platform. xAI plans to gradually release the model to a wider audience and introduce new features in the coming days.
What you should know about Grok?
Grok-1.5 is the successor to Grok-1, which was released in November 2023 as an AI assistant modelled after the Hitchhiker’s Guide to the Galaxy, designed to answer questions and offer insights.
Grok, according to xAI, is designed to answer questions with a “bit of wit and a rebellious streak.” The AI tool answers questions using real time knowledge of the world from the X platform. To create Grok, xAI built a custom training and interference stack based on Kubernates, Rust and JAX.
xAI said its mission is to develop AI tools that empower humanity’s pursuit of knowledge and understanding. Grok-1.5 is a significant step towards this goal, offering researchers and users a powerful tool for various tasks, according to the company.
Home
