Late Monday, xAI, the artificial intelligence (AI) venture helmed by Elon Musk, announced its highly anticipated Grok 3 model, accompanied by enhanced features for its iOS and web applications.
The release marks a notable step forward for xAI’s flagship chatbot, the company’s direct competitor to established AI powerhouses like OpenAI’s GPT-4 and Google’s Gemini. It also comes at an opportune time with new rivals such as DeepSeek gaining prominence lately.
Grok’s integration into Musk’s social media platform X enables various features, including image analysis and chatting capabilities like ChatGPT and others. While xAI had ambitiously targeted a 2024 release for Grok 3 after months of intensive development, the company ultimately delivered it later than planned.
Grok 3 was trained from a massive Memphis-based data center with approximately 200,000 GPUs. According to Elon Musk’s recent statement on X, Grok 3’s development used computing resources ten times more powerful than those for its previous iteration, Grok 2. The expanded training dataset also includes court case documentation.
The Grok 3 release introduces a suite of AI models, including a streamlined variant called Grok 3 mini, which prioritizes rapid response times while accepting a modest trade-off in accuracy. While Monday marked the initial phase of the rollout, the complete range of Grok 3 models and their associated features are not fully available as of yet, with several still undergoing beta testing.
According to xAI, Grok 3 has shown superior performance than GPT-4o across multiple benchmark tests, including AIME’s mathematical assessment and GPQA’s evaluation of doctoral-level scientific problems spanning physics, biology, and chemistry. The company also reports that an early iteration of Grok 3 achieved notable results in Chatbot Arena, a public testing platform where performance ratings are individual user-based.
The Grok 3 lineup features two specialized variants, Grok 3 Reasoning and Grok 3 mini Reasoning, engineered with advanced problem-solving capabilities similar to those found in OpenAI’s o3-mini and DeepSeek’s R1 models. These reasoning-focused versions include self-verification to reduce common AI errors through fact-checking processes.
In a direct comparison with OpenAI’s most advanced mini model, o3-mini-high, xAI asserts that Grok 3 Reasoning has achieved higher scores across various industry benchmarks, including the recently introduced AIME 2025 mathematics evaluation standard.
Through the Grok app, users can access these reasoning models, with two modes at their disposal. The standard “Think” command activates Grok 3’s basic reasoning capabilities, while a more resource-intensive “Big Brain” mode can be engaged for tackling complex problems requiring enhanced computational power.
Stay Connected with ProPakistani
Get the latest tech news, telecom insights, and product launches wherever you prefer.
Add ProPakistani to Preferred Sources and see more of our stories in Google Search and Top Stories.
