New Grok 3 Debuts with Unconventional Approach
Elon Musk‘s artificial intelligence venture, xAI, has officially launched Grok 3, its most advanced AI model yet. This latest iteration marks a significant step forward in AI reasoning, problem-solving, and instruction-following capabilities. The latest AI model has been trained on xAI’s Colossus supercomputer, which provides 10 times the computational power of its predecessor. This increase in computing capacity has resulted in notable improvements in reasoning, mathematics, coding, and complex decision-making.
Enhanced Problem-Solving with “Think” Mode
A standout feature of Grok 3 is its enhanced reasoning ability, developed through large-scale reinforcement learning. The model can engage in extended thought processes lasting seconds to minutes, allowing it to:
- Correct its own mistakes
- Explore alternative solutions
- Provide more accurate and well-structured answers
Users can access this capability through the new “Think” mode, which offers a transparent look at the AI’s decision-making process.
Benchmark Performance: Setting New Standards
The new AI model has excelled in key benchmark tests, showcasing its superior reasoning and computational abilities:
- 93.3% on the 2025 American Invitational Mathematics Examination (AIME)
- 84.6% on graduate-level expert reasoning assessments
- 79.4% on LiveCodeBench for coding and problem-solving
Introducing Grok 3 Mini: A Cost-Efficient Alternative
To meet diverse user needs, xAI has also introduced Grok 3 Mini—a lightweight, cost-effective version optimised for reasoning-focused tasks. Despite being a smaller model, it delivers impressive performance:
- 95.8% on AIME 2024
- 80.4% on LiveCodeBench
This makes the Mini a powerful yet efficient option for users who don’t require the full extent of world knowledge processing.
Rollout and Availability
Grok 3 is now being gradually rolled out to users, with an early preview of its enhanced reasoning abilities. This launch cements xAI’s position at the forefront of AI development, providing cutting-edge tools for complex problem-solving and advanced decision-making.