On Wednesday night, Elon Musk unveiled the latest version of xAI’s flagship AI model, Grok 4, in a livestream that showcased its top-tier performance on AI benchmarks.

Why it matters: xAI is locked in a fast-paced race against OpenAI, Anthropic, and other players to deliver faster, more reliable AI—meeting the soaring expectations of investors.

Key highlights: Musk and xAI researchers demonstrated Grok 4 solving an advanced math problem, generating an image of two colliding black holes, and even predicting the most likely winner of next year’s World Series (the Dodgers, with a 21.6% chance).

The company claims Grok 4 now scores highest on Humanity’s Last Exam—a set of highly challenging benchmarks—as well as other key measures of AI progress.

Under the hood: Grok 4 includes a new "heavy" mode, deploying multiple AI agents to collaborate on solving particularly tough problems. The model is available now, with the advanced tier priced at $300/month.

Researchers also demoed a new voice mode, more natural and responsive than before—though it stumbled when asked to compose an "opera about Diet Coke," producing something closer to a singsong Shakespearean monologue.

Musk’s take: The xAI CEO expressed amazement at Grok’s rapid advancements, declaring: "Grok 4 is smarter than nearly all graduate students across all disciplines simultaneously."

The launch intensifies the race toward artificial general intelligence (AGI), as xAI cements its place among the industry’s top contenders.