Grok 3, launched by xAI on February 17, 2025, claims to be the “world’s smartest AI.” Powered by 200,000 Nvidia H100 GPUs, it scores 93.3% on the 2025 AIME and 1402 Elo in Chatbot Arena. With DeepSearch and Think mode, it’s free on X (10-message limit) or Premium+ ($22/month). In beta, it’s growing, with voice and API soon.
Grok 3 outclasses Grok 1 and 2 with 10x compute and a 1-million-token context. Think mode boosts reasoning beyond earlier models’ chat focus, while DeepSearch adds real-time data. It hits 84.6% on science and 79.4% on coding, with mini and multimodal perks.
Grok 3’s Aurora-powered image generation beats Grok 1’s text-only and Grok 2’s basics. It offers photorealism and less censorship but lags behind DALL·E 3 or MidJourney in Polish. Integrated with Grok’s tools, it’s handy on X (10-image limit for free users).
TLDR: Grok 3’s scale, reasoning, and features, like DeepSearch, outshines its predecessors. Aurora adds decent visuals, though not top-tier. Free with limits or premium, it’s a beta star for math, code, and quick images.
RELATED STORIES:
TAKE ACTION:
