What is Grok 4
Grok 4 is the latest AI large model launched by xAI. Grok 4's reasoning ability is 10 times improved compared to the previous generation. The model has excellent reasoning ability, can approach full marks in highly difficult exams such as the SAT and GRE, and surpasses other frontier models in multiple benchmark tests. Grok 4 supports multimodal functions, can understand subjective concepts, generate code and visual content, and has made major improvements in voice interaction. Grok 4 is divided into two versions. Grok 4 is the single agent version, and Grok 4 Heavy is the multi agents version, supporting four agents working simultaneously, with a context window supporting up to 256k tokens.
Main features of Grok 4
- Excellent reasoning ability: Approaches full marks in highly difficult exams such as the SAT and GRE, demonstrating a level of reasoning beyond humans.
- Multimodal understanding: Can understand subjective concepts, and search and analyze images.
- Information integration and summarization: Integrates information from channels such as social media, extracts key events, and sorts them chronologically.
- Code and visualization generation: Generates complex animations based on scientific prompts, such as simulating black hole collisions.
- Improved voice interaction: Supports five new voices, with smoother conversations and more natural emotional expression.
- Complex task handling: Performs excellently in complex tasks such as simulated business management, with strong strategic planning and execution capabilities.
- Parallel agent collaboration: The SuperGrok Heavy version supports multiple agents solving complex problems in parallel.
Grok 4's test performance
- Official testing：
- Humanity’s Last Exam: Contains 2,500 interdisciplinary expert-level questions. Grok 4 Heavy scored 44.4% when using tools, and with further optimization it can be increased to 50.7%.
- AIME25 (math competition): Grok 4 Heavy achieved a perfect score of 100%, crushing other models.
- GPQA (graduate-level question answering): Grok 4 Heavy scored 88.9%, ahead of Gemini 2.5 Pro (86.4%) and Claude 4 Opus (79.6%).
- HMMT25 (high school math competition): Grok 4 Heavy scored 96.7%, far exceeding Gemini 2.5 Pro (82.5%).
- USAMO25 (USA Mathematical Olympiad): Grok 4 Heavy scored 61.9%, far ahead of Gemini DeepThink (49.4%) and Gemini 2.5 Pro (34.5%).
- ARC-AGI (abstract reasoning): Grok 4 scored 15.9%, nearly double the previous commercial SOTA.
- Vending-Bench (simulated business management): Grok 4 earned a net profit of $4694, far exceeding Claude Opus 4 ($2077) and human players ($844).
- Third-party evaluation(Tested by the large model performance evaluation platform Artificial Analysis):
- Artificial Intelligence Index: Grok 4 scored 73 points, surpassing OpenAI o3 (70 points), Google Gemini 2.5 Pro (70 points), Anthropic Claude 4 Opus (64 points), and DeepSeek R1 0528 (68 points).
- Coding Index and Math Index: Grok 4 ranks first in both.
- GPQA Diamond score: Reached a record high of 88%, exceeding Gemini 2.5 Pro's 84%.
- Humanity’s Last Exam score: A record high, reaching 24%, surpassing Gemini 2.5 Pro's 21%.
- Speed: Grok 4 is 75 tokens/second, though not as fast as o3 (188 tokens/second) and Gemini 2.5 Pro (142 tokens/second), it is better than Claude 4 Opus Thinking (66 tokens/second).
Grok 4 product pricing
- Paid subscription plans：
- SuperGrok: $300 per year, $30 per month.
- SuperGrok Heavy: $3,000 per year, $300 per month.
- API call pricing：
- Input: $3 per million tokens.
- Output: $15 per million tokens.
Grok 4 official website address
- Official website address：Grok
Grok 4 application scenarios
- Educational tutoring: Provides students with personalized learning plans, answers complex academic questions, and helps students better understand and master knowledge.
- Scientific research: Able to analyze large amounts of experimental data, predict scientific trends, and help scientists discover new theories and technologies.
- Business and finance: Conducts market analysis and forecasting, provides data support for the formulation of corporate business strategies, and optimizes corporate operational efficiency.
- Content creation: Assists creative generation, supports scriptwriting and animation production in fields such as advertising, film and television, and gaming, improving creative efficiency.
- Intelligent assistant: As an intelligent voice assistant, processes multimodal information, helps users complete daily tasks, and improves convenience in life.