Just a few days after its debut, xAI’s Grok-3 AI chatbot has transitioned from a paid service costing $40 per month to offering free access, though only for a limited period. Elon Musk, head of xAI, has indicated that this complimentary access will last only a “short time,” leaving users uncertain about how long this opportunity will be available.
At present, users can experiment with two key features: Think and DeepSearch. The Think feature enhances Grok-3 by incorporating reasoning abilities into its interactions, comparable to systems like DeepThink on DeepSeek, Google’s Gemini 2.0 Flash Thinking Experimental, and OpenAI’s various models.
In contrast, DeepSearch serves as xAI’s counterpart to the Deep Research capabilities offered by platforms like Perplexity, Gemini, and ChatGPT. Grok-3 marks the second widely used AI tool to provide free access to such complex processes as DeepSearch or Deep Research.
This functionality represents an exciting advancement in the field of AI as it revolutionizes the knowledge-gathering process. When users submit a question, the AI dissects it and formulates a research strategy, identifying specific sources from which information will be sourced.
As the AI navigates through various knowledge repositories, it analyzes the data in real-time and delivers findings in the form of an in-depth report.
To access Grok-3 on mobile, visit the X app and select the Grok icon located in the navigation bar. Alternatively, you can access it directly on the web by visiting the x.com/i/grok page.
There is substance to all the hype
In my early interactions with DeepSearch, I’ve found it quite impressive. Unlike other platforms that require approval for their research methods, Grok-3 immediately begins the task upon question submission, and it does so with remarkable speed.
For instance, when I queried the scientific research related to screen time’s impact on younger audiences, Grok-3 produced a report notably faster than Gemini. Additionally, it offered greater transparency by allowing me to observe its reasoning process while compiling answers.
Moreover, it provides a detailed breakdown of the entire process, which is saved as its dataset, displayed above the final answer. However, one limitation is that users cannot modify the research plan prior to initiating the search.
In contrast, Gemini allows for research plan adjustments but lacks the transparency of a breakdown or clear reasoning process. Gemini’s results referenced material from 37 different sources, providing extensive citations, while Grok-3’s DeepSearch succinctly summarized its conclusions with only six key references, yet the information was equally valuable.
When exploring the less complex Think feature, Grok-3 again outperformed Gemini in terms of speed. Upon inquiring about the implications of Microsoft’s new quantum computing chip, I noted significant differences in approach.
Grok-3 approached the question holistically, addressing both the scientific applications and benefits, along with the associated risks, dedicating a section solely to this. Furthermore, it displayed its reasoning at all times throughout the interaction.
Conversely, Gemini appeared more reserved and academically focused, whereas Grok-3 articulated complex ideas in a more accessible manner. While it’s difficult to determine a definitive leader among these AI models, benchmarks provided by xAI suggest that Grok-3 has outperformed models from Google, OpenAI, DeepSeek, and Anthropic across various evaluations.
