Elon Musk’s artificial intelligence venture, xAI, has rolled out Grok-1.5, an upgraded version of its Grok AI model. The company claims Grok-1.5 demonstrates significant advancements in mathematical problem-solving and coding abilities. xAI’s internal testing suggests Grok-1.5 may even be competitive with cutting-edge AI models like GPT-4 and Claude 3 Opus.
Launched initially in November 2023, Grok began as an ambitious project by Musk’s xAI, aiming to differentiate itself in the crowded field of generative AI. Grok-1, the first significant iteration, boasted an ability to understand and generate code, achieving a 63.2% success rate on the HumanEval coding task and a 73% success rate on the MMLU, a multidisciplinary multiple choice test. This marked a significant advancement from its predecessor, Grok-0, a prototype with 33 billion parameters.
Musk, always eager to challenge the AI status quo, positions Grok-1.5 as a potential rival to the popular ChatGPT chatbot from OpenAI. xAI is particularly focused on Grok’s mathematical competency, demonstrated by significant benchmark gains. Grok-1.5 reportedly scored over 50% on the MATH benchmark, more than double the capability of its predecessor. Additionally, the new model excelled on coding-related tests like GSM8K and HumanEval, showcasing promise in understanding and generating programming languages.
Also Read: Elon Musk Open-Sources Grok AI, But Training Data Remains Secret
One of Grok-1.5’s key improvements is a vastly expanded context window. It can process up to 128,000 tokens, allowing the AI to draw on significantly more information to understand complex situations and instructions. This grants Grok a memory capacity up to 16 times larger than the previous model. However, xAI hasn’t released details on Grok’s potential improvements outside of math and coding, leaving analysts curious about its performance in other areas.
The race to create the most powerful AI language model is heating up. ChatGPT 5, anticipated this summer, promises a human-like conversational experience. It remains to be seen if Grok-1.5 will maintain its initial competitive edge.
Currently, Grok is restricted to users of X’s (formerly Twitter) Premium+ subscription tier, though Elon Musk has promised to widen availability to standard Premium users in the future.
