Today, Elon Musk's xAI officially launched its latest model, Grok3, which Musk claims is currently the "smartest AI on Earth." It will compete with ChatGPT, Gemini, DeepSeek, and Claude.
During the launch event, xAI showcased many benchmark tests, where Grok3 outperformed Gemini2Pro, Claude3.5Sonnet, GPT4o, and DeepSeek V3 in mathematics, science, and coding benchmarks.
Elon Musk and his team revealed that Grok3 will also have reasoning capabilities similar to DeepSeek R1 and OpenAI o3Mini. Like DeepSeek R1 and the current ChatGPT o3Mini, Grok3 will display detailed thought processes, allowing users to see how the chatbot approaches problems. Musk confirmed that xAI would slightly obscure its reasoning so that the company cannot directly replicate the chatbot.
It is understood that Grok3 also features Deep Search capabilities. An xAI employee explained: "It is our first-generation comprehensive agent that not only helps engineers, researchers, and scientists with coding but can actually assist anyone in answering the questions they encounter daily."
Regarding search capabilities, Musk stated that Grok3's deep search will save a significant amount of time. For example, if it used to take half an hour or an hour to search online or on social media, you can just let Grok3 handle it, and ten minutes later, it will have completed an hour's worth of work for you.
As for the Grok voice assistant, xAI indicated that the Grok voice assistant is currently being refined and will be available to users in the coming weeks.
Notably, an xAI staff member asked when Grok would be deployed on SpaceX rockets, to which Musk replied: "In two years."
Officially, Grok3 will first support Premium Plus subscribers on x. Therefore, this is the initial group to gain access, and to see all premium features, the x application is required.
Additionally, Grok3 will be available on the official website and iOS application. Musk emphasized that it is currently more like a beta version, which means there may be some imperfections at the start, but the official team will quickly improve it, with updates almost daily.