Musk's Grok 4 has been released and has reached the peak of AI, scoring 73 points in the comprehensive benchmark tests to take the top spot, surpassing OpenAI's o3 (70 points), Google's Gemini 2.5 Pro (70 points), Anthropic's Claude 4 Opus (64 points), and DeepSeek R1 0528 (68 points), becoming the most advanced AI model currently available. This is the first time xAI has taken a leading position in the forefront of AI.
Grok 4 has excelled in key metrics such as coding (LiveCodeBench & SciCode) and mathematics (AIME24 & MATH-500), particularly setting historical high scores in the GPQA Diamond and the final human exam HLE tests (88% and 24% respectively), where previous leaked versions of HLE claimed to achieve 45%, clearly an overestimation of Grok 4. Its context window is 256k tokens, supporting text and image inputs as well as function calls, with pricing identical to Grok 3, but slightly slower than competitors.