xAI Unveils Updated Chatbot Grok-1.5

Elon Musk’s company xAI has announced an updated version of its large language model, Grok-1.5.

According to the press release, the chatbot’s capabilities will approach the performance of both open and closed LLM, including OpenAI’s GPT-4 and Anthropic’s Claude 3.

Elon Musk noted that Grok-1.5 will be available on the X platform. The next neural network model is already in training. According to him, it will be able to “exceed current AI on all metrics,” though he did not specify a release date.

Should be available on ? next week.

Grok 2 should exceed current AI on all metrics. In training now. https://t.co/s7rVxvK8mZ

— Elon Musk (@elonmusk) March 29, 2024

The company expects Grok-1.5 to outperform the previous model in many tests, including programming and mathematics.

“Grok-1.5 scored 50.6% in MATH and 90% in GSM8K — two mathematical benchmarks covering a wide range of competitive tasks for schoolchildren and high school students. Additionally, it scored 74.1% in HumanEval, which assesses code generation and problem-solving abilities,” xAI noted in its blog.

In the MMLU test, which evaluates AI models’ language understanding capabilities for various tasks, the new model scored 81.3%, significantly surpassing its predecessor’s 73%.

xAI also reported that Grok-1.5 has a context window of up to 128,000 tokens. This allows the model to perceive and process large volumes of information simultaneously — 16 times more than Grok-1. This makes it suitable for analyzing, summarizing, and extracting information from lengthy documents.

According to the company, the new AI model will initially be available to testers and those already using Grok on the X platform. Deployment will occur in stages: the company will enhance the chatbot and introduce several new features. Gradually, it will become accessible to a broader range of users.

Back in March, xAI opened Grok’s source code after Elon Musk accused OpenAI of a non-public approach to developing ChatGPT.