xAI launches Grok 3 to compete with ChatGPT, DeepSeek，Gemini
Elon Musk's company, xAI, has released its latest AI model, Grok 3, which shows significant improvements over Grok 2.
In a short period, Grok 3 has achieved functionalities an order of magnitude more powerful than Grok 2. Compared to Grok 2, Grok 3's training data has increased tenfold, outperforming or rivaling competitors like ChatGPT, DeepSeek, and Gemini. Although Grok started later, it quickly caught up to ChatGPT in MMLU scores.
Elon Musk Calls Grok 3 the Smartest AI on Earth
After the official launch of the latest Grok 3 model, Elon Musk referred to it as the "smartest AI on Earth." Why does he make such bold claims? Let's take a look at the test scores released by xAI:
- Grok 3 and Grok 3 mini excelled in mathematical, scientific, and programming benchmark tests, surpassing all mainstream models, including GPT-4o, Claude 3.5 Sonnet, DeepSeek-V3, and Gemini-2 Pro.
- The reasoning capability of Grok 3, specifically the Grok-3 Reasoning Beta and Grok-3 mini Reasoning, also performed exceptionally well in mathematical, scientific, and programming benchmark tests, outpacing similar reasoning models like DeepSeek-R1 and OpenAI's o3 mini.
The results from the tests indicate that Grok 3 and Grok 3 mini performed impressively across multiple benchmarks, surpassing several mainstream models, including GPT-4o, Claude 3.5 Sonnet, DeepSeek-V3, and Gemini-2 Pro.
xAI Launches Grok 3 Intelligent Search Engine DeepSearch
In addition to Grok-3, xAI has also launched an intelligent search engine called DeepSearch, which works in collaboration with Grok-3. This is xAI’s first generation of a wide-ranging agent tool that not only assists engineers, researchers, and scientists in coding but also helps everyone answer everyday questions.
DeepSearch can not only search the web and find existing information, but it can also infer the user’s true intentions and think critically. After cross-referencing various information sources, it ensures that the correct information is returned.
Let's take a look at Musk's demonstration: “The timing of the next Starship launch.”
DeepSearch can think critically and then search through web news, X platform posts, etc., to analyze and summarize the information before giving a response. Users can view the model's reasoning process and it supports tasks such as research and data analysis.
Models like DeepSeek and OpenAI's ChatGPT have already ventured into AI search engines, providing real-time analysis and answers via the internet. It's clear that AI-powered online search is becoming a popular area of business for major AI companies.
What You Need to Know About Grok 3
Grok 3 will first be available to Premium+ subscribers on X, who will be among the first to access it. It is unclear when regular users will be able to use it.
Because the model is still being updated and improved, the version launched on the Apple App Store will lag behind, whereas the web version will receive the most timely updates. It has been suggested that a polished version may be launched in about a week. The xAI team has also indicated plans to develop voice interaction features for Grok 3, which could be one of the best experiences of using Grok 3.
Musk has confirmed plans to open-source Grok 3, releasing older versions as new ones are developed. The model's gender has not yet been determined.
Competition Between Grok 3, ChatGPT, DeepSeek, and Gemini
In the increasingly competitive global AI landscape, the launch of Grok 3 has garnered significant industry attention. Particularly as the new Chinese company DeepSeek showcased its cost-effective model, which rivals the performance of OpenAI's GPT and o1/o3 series, it has prompted many AI companies to rethink whether simply relying on increased computing resources and model size is still the optimal solution.
In early February, Google released a series of new models, including Gemini 2.0 Flash, Gemini 2.0 Flash-Lite, and the experimental version of the new flagship large model Gemini 2.0 Pro.
On the same day Grok 3 was launched, Ultraman also mentioned on X, “For testers with high expectations, the experience from trying GPT-4.5 to feel AGI is far deeper than I anticipated!” This seemingly signals that GPT-4.5 is undergoing testing and may not be far from its official release.
Anthropic has also announced plans to launch the Claude 4 series soon.
As OpenAI, Anthropic, Meta, and Google continuously release more advanced AI models, the AI battle is expected to escalate further. Whether Grok 3 can truly surpass its competitors and become the smartest AI on Earth remains to be seen as the market continues to evolve.
XXAI Continuously Optimizing Amid Competition
As Grok, ChatGPT, DeepSeek, and Gemini models continue to upgrade and advance, XXAI is continuously updating and adding models, aiming to provide users with access to all models' services on a single platform.
XXAI plans to launch a web version this month and will also add features such as XXAI bot, XXAI tools, and multi-model collaboration on top of the existing XXAI chat functionality. Interested users can look forward to these developments.