Grok 4 User Guide AIME full score
Musk xAI "PhD-level" AI, the most powerful model on Earth
๐ Quick Overview
๐ Release time
July 9, 2025
๐ข Development Company
xAI (founded by Musk)
๐ง Context
256K tokens (multimodal)
๐ฐ Pricing
SuperGrok $30/month or $3000/year
๐ Grok development history
Grok 1 Posted
First generation model, humorous conversation style
Grok 2 released
Performance improvement, support image generation
Grok 3 released
Thought chain reasoning, 100,000 H100 training
Grok 4 + Heavy released ๐ฅ
Skip 3.5, AIME 2025 full score, PhD level ability
๐ค Grok 4 model family
โก Grok 4 Ultimate version
All-round multi-modal model, PhD level capabilities
- Context: 256K tokens
- Modal: Text + Image input
- Capabilities: Graduate/PhD level performance
- Features: Thought chain reasoning, real-time X data
๐ Grok 4 Heavy The Strongest Agent
Multi-agent system, full score in AIME 2025
- Architecture: Multi-Agent collaboration system
- AIME 2025๏ผ Full score (American Mathematics Invitational Competition)
- Beyond: All existing large models
- Pricing: SuperGrok exclusive
๐ Grok 4 Fast Cost-effective version
1/25 The cost reaches Gemini 2.5 Pro level
- Positioning: Cost-effective Agent application
- Capabilities: Independent browsing, multiple tool chain tasks
- Cost: Very low
โจ Grok 4 Core Competencies
๐ง Doctoral Reasoning
Musk said that "all aspects have reached the level of graduate students or even doctoral students"
๐ข Full marks in mathematics
AIME 2025 full score, surpassing all competing products
๐ป Coding ability
Image generation/editing (2025.8), video (2025.9-10)
๐ Real-time search
X/Twitter platform real-time data access
๐ค Multiple Agents
Grok 4 Heavy multi-agent collaboration
๐ Humor style
Rebellious and interesting, dare to discuss sensitive topics
๐ Performance comparison (December 2025)
| Test | Grok 4 Heavy | GPT-5.2 Pro | Gemini 3 Pro | Claude Opus 4.5 |
|---|---|---|---|---|
| AIME 2025 | 100% perfect score | ~85% | ~80% | ~75% |
| Mathematical Reasoning | Strongest | strong | strong | strong |
| Agent capabilities | Multiple Agents | Single Agent | Antigravity | Computer Use |
| Real-time data | X platform | Internet search | Google Search | โ |
| Context | 256K | ~400K | 1 million | 200K |
๐ฐ Pricing Plan
X Premium+
- Grok 4 Basic Access
- Usage restrictions apply
- X platform privileges
SuperGrok
- Grok 4 Heavy Full Access
- Unlimited use
- Priority access to new features
- API access
๐ How to use
๐ Web version
๐ฑ X App
Sidebar Grok Icon
๐ฒ Standalone App
iOS/Android Grok App
๐ API
x.ai Developer Platform
๐ก Recommended usage scenarios
๐ข Mathematics Competition
AIME full score level mathematical reasoning
๐ Real-time analysis
X platform hot spots and public opinion analysis
๐ค Complex Agent
Multi-step automation task (Heavy)
๐ Creative Dialogue
Humorous style, no excessive restrictions
โ FAQ
What is the difference between Grok 4 Heavy and regular Grok 4?
Heavy is a multi-agent system suitable for complex reasoning tasks and performs better in mathematical tests such as AIME. The regular version is suitable for daily use.
Is $3000/year worth it?
Worth it for users who need top-level mathematical reasoning, research-level tasks. Premium+ at $16/month for regular users is enough.
Does Grok 4 support Chinese?
Supported, but English is better.
Why was Grok 3.5 skipped?
xAI jumps directly from Grok 3 to Grok 4, demonstrating its determination to accelerate in the AI race.