0:00 Introduction: Grok 3 is presented as the world's smartest AI, surpassing state-of-the-art models on various benchmarks.
0:00 Benchmarks: Grok 3 and Grok 3 Mini excel on benchmarks, outperforming Gemini 2, DeepSeek V3, Claude 3.5 Sonnet, and GPT-4o.
0:02 Chatbot Arena: Grok 3 is the number one model in the Chatbot Arena, a blind test where users select the better AI response without knowing the model's identity.
0:03 Blind Test: Grok 3 reached an ELO score of 1400 in a blind test, indicating top performance across categories like instruction following and coding.
0:04 Continuous Improvement: Grok 3 is continuously improving, with users potentially noticing enhancements almost daily.
0:04 Reasoning Models: Grok 3's reasoning capabilities, where models think for an extended period, also surpass previous models.
0:05 Beta Version: A beta version of the Grok 3 reasoning model is being tested, along with a mini version trained for a longer duration.
0:06 Benchmarks Categories: Grok 3 is evaluated on mathematics (high school competition problems), science (PhD-level questions), and coding (competitive coding, LeetCode).
0:07 Test Compute: Grok models can spend more time reasoning before providing an answer, improving performance.
0:08 Generalization Testing: Grok 3 was tested on the new AIME 2025, showing strong generalization capabilities compared to the smaller model.
0:10 Demonstration: A demonstration showcases Grok's ability to generate code for an animated 3D plot of a launch from Earth to Mars and back.
0:13 Deep Search: A new product called Deep Search, the first generation of Grok agents, is introduced to help engineers, researchers, and everyone else answer questions.
0:14 Capabilities of Deep Search: Deep Search acts as a next-generation search engine that deeply analyzes user intent, considers relevant facts, and cross-validates different sources.
0:16 Chain of Thought Transparency: Users can examine the model's chain of thought in Deep Search to understand its reasoning process.
0:17 Rollout and Availability: Grok will be accessible on the new website grok.com and through the Grok app in the iOS App Store.
0:18 Version Differences: The web version on grok.com will be the latest and most advanced due to app store approval processes and phone format limitations.