Xai unveiled the Grok 3 AI model on Monday, along with new features such as image analysis and sophisticated question answering.
The company has used an immeasurable data center with around 200,000 GPUs to develop the GROK 3. According to Xai owner Elon Musk, the project is reported to contain information from the law, leveraging “10 times” of computing power than its predecessor, the Grok 2. Case filing.
Musk argued that Grok 3 is “an AI seeking the greatest truth,” even if politically modified ones and its truth sometimes opposed.
The Grok 3 rollout includes a family of models designed for a variety of needs. For example, the Grok 3 Mini prioritizes response times over absolute accuracy. Of particular note, however, is the GROK 3 model, which focuses on new reasons.
These variations, called Grok 3 Reasoning and Grok 3 Mini Reasoning, aim to emulate human-like cognitive processes by “thinking” the problem. Comparable to models such as Openai’s O3-Mini and Deepseek’s R1, these inference systems attempt to fact-check the response by reducing the likelihood of errors and missteps.
Grok 3: Benchmark Results
Xai claims that the Grok 3 outweighs Openai’s GPT-4O in certain benchmarks, including AIME and GPQA. It assesses the proficiency of models tackling complex problems in mathematics, physics, biology and chemistry.
The early version of the Grok 3 currently leads the Chatbot Arena. It is a crowdsourced evaluation platform where users model AI against each other and rank the output. This model was the first to break Arena’s 1400 score.
According to Xai, Grok 3 reasoning is superior to its rivals in various prominent benchmarks.
These inference models are already integrated into features available from the Grok app. Users can select commands such as “Think” or activate a more computationally intensive “big brain” mode to address particularly challenging questions.
Xai positions inference models as the ideal tool for STEM (science, technology, engineering, mathematics) applications, including mathematics, science and coding challenges.
Protecting against AI distillation
Interestingly, not all of Grok 3’s internal processes are exposed to the user. Musk explained that some of the “thinking” in the inference model are intentionally obscure to prevent distillation. This is a controversial practice in which competing AI developers extract knowledge from their own models.
The practice has entered the spotlight in recent weeks after Chinese AI company Deepseek faced allegations that it would distill Openai’s model and develop its latest model, the R-1.
Xai’s new inference model serves as the foundation for a new GROK app feature called DeepSearch. This feature uses the GROK model to scan the Internet and Musk social platform X for related information before integrating detailed summary in answers to user queries.
Go to Grok 3 and commit to open source
Currently, access to the latest GROK models is associated with the subscription tier of X. Premium+ subscribers who pay $50 per month (~£41) will receive priority access to the latest features.
Xai also introduces a SuperGrok subscription plan, which is reportedly priced at $30 a month or $300 a year. SuperGrok subscribers benefit from enhanced inference, deeper search queries, and unlimited image generation capabilities.
The company also teased future features. Within a week, the Grok app will introduce a voice mode that allows you to interact with AI via synthetic voice similar to Gemini Live.
Musk has also revealed plans to release the GROK 3 model via an enterprise-ready API in the coming weeks, including the DeepSearch feature.
Grok 3 is still fresh, but Xai plans to open source its predecessor of open source in the coming months. Musk claims that Xai will continue to sell the last version of Grok in open source.
“If Grok 3 is mature and stable, and if it’s probably within a few months, I’ll explain Grok 2 in Open Source 2,” explains Musk.
“Wake Up” AI Model
For a long time, Grok has been unfiltered, bold and willing to sell for questions that competitors might avoid. Musk previously described AI as a “conflict” and presented it as a model that is not afraid to be exposed to controversial topics.
True to that promise, early models like Grok and Grok 2 accepted politically charged queries, even directing them towards colorful languages when prompted. However, these versions also revealed some biases when delving deeper into political discourse.
“We’re working to bring Groke closer to politically neutral,” Musk said.
However, it remains to be seen whether Grok 3 will achieve this goal. With these changes, analysts have already emphasized the social impact of increasingly “seeking for the truth” introducing politically sensitive AI systems.
At Grok 3, Musk and Xai issued bold statements to advance the technology while potentially promoting debate on bias, transparency and the ethics of AI deployment.
As competitors like Openai, Google, and Deepseek come to offer, Grok 3’s success depends on their ability to balance accuracy, user demand and social responsibility.
Reference: AI in 2025: Purpose-driven models, human integration, etc.
Want to learn more about AI and big data from industry leaders? Check out the AI & Big Data Expo in Amsterdam, California and London. The comprehensive event will be held in collaboration with other major events, including the Intelligent Automation Conference, Blockx, Digital Transformation Week, and Cyber Security & Cloud Expo.
Check out other upcoming Enterprise Technology events and webinars with TechForge here.
