Elon Musk’s AI venture, xAI, has launched Grok 3, the latest iteration of its flagship AI model. This ambitious release introduces enhanced capabilities for the Grok app and web platform.
Grok 3, comparable to OpenAI’s GPT-4o and Google’s Gemini, possesses the ability to analyze images, answer complex questions, and drive various features on X, Musk’s social network. Trained on a massive dataset using 200,000 GPUs, Grok 3 boasts 10 times the computational power of its predecessor, Grok 2.
Musk highlights Grok 3 as an “order of magnitude more capable” AI, prioritizing truthfulness even if it challenges societal norms. The model comprises a family of variations, including Grok 3 mini, which offers faster responses at a slight compromise in accuracy.
xAI claims that Grok 3 outperforms competitors like GPT-4o in benchmarks such as AIME and GPQA. Additionally, it incorporates “reasoning” capabilities, similar to OpenAI’s o3-mini and DeepSeek’s R1, which improve fact-checking and accuracy.
The Grok app now features reasoning models accessible through “Think” and “Big Brain” modes for more challenging queries. These models excel in mathematical, scientific, and programming domains. Additionally, the new DeepSearch feature allows Grok to analyze information from the internet and X, providing comprehensive abstracts.
xAI has introduced tiered subscription plans: Premium+ ($50/month) grants early access to Grok 3, while the upcoming SuperGrok ($30/month or $300/year) unlocks additional reasoning and DeepSearch queries, as well as unlimited image generation.
Voice integration is expected within a week, and Grok 3 models will be made available through xAI’s enterprise API alongside DeepSearch capabilities. Notably, xAI plans to open-source Grok 2 in the near future.
Grok’s initial launch positioned it as an unfiltered AI, willing to address controversial topics. However, it faced criticism for its left-leaning bias on political issues. Musk has emphasized the influence of training data on Grok’s behavior and aims to shift the model towards political neutrality. The full extent of these adjustments and their potential implications remain to be seen.
Original source: Read the full article on TechCrunch
