# GPT-5 just caught them *all*  (Grok 4.20 and Gemini 3.0)

Source: https://www.youtube.com/watch?v=ikBODkpu7Jo
Recap page: https://rapidrecap.app/video/ikBODkpu7Jo
Generated: 2025-08-16T03:02:36.906+00:00

---
## Quick Overview

OpenAI's GPT-5 achieved gold medal-level performance in the 2025 International Olympiad on Informatics (IOI), ranking first among AI participants and sixth relative to humans, demonstrating significant advancements in AI reasoning capabilities without specific IOI training.

**Key Points:**
- OpenAI's GPT-5 achieved gold medal-level performance in the 2025 International Olympiad on Informatics (IOI), ranking first among AI participants and sixth overall.
- The GPT-5 model demonstrated advanced general-purpose reasoning capabilities without specific IOI training.
- Elon Musk is moving forward with Grok 4.20 and has left xAI to pursue a new venture.
- Leo Aschenbrenner, a former OpenAI researcher, has launched an AI-focused hedge fund named "Situational Awareness" that manages over $1.5 billion.
- Google AI has released new Imegen 4 and Gemma 3 models, indicating ongoing development in AI image generation and efficient language models.
- The concept of "frontier models" becoming more agentic and capable is discussed, emphasizing the critical need for AI safety research.
- OpenAI's reasoning system also achieved a high ranking in a programming competition, showcasing its versatility.

![Screenshot at 11:41: Greg Brockman announces that an OpenAI reasoning system achieved gold medal-level performance \(#6 relative to humans, #1 relative to AIs\) in the International Olympiad on Informatics \(IOI\) without specific training.](https://ss.rapidrecap.app/screens/ikBODkpu7Jo/00-11-41.png)

**Context:** The video discusses recent advancements and news in the AI field, focusing on the performance of major AI models like GPT-5 and the strategic moves of key figures in the industry. It highlights achievements in AI reasoning and the growing investment in AI-focused ventures.

## Detailed Analysis

This video discusses the impressive performance of OpenAI's GPT-5 model in the 2025 International Olympiad on Informatics (IOI). Greg Brockman and Sheryl Hsu both highlight that GPT-5 achieved a gold medal-level performance, ranking first among AI participants and sixth overall when compared to human competitors. Notably, this was accomplished without any IOI-specific training, showcasing the model's advanced general-purpose reasoning abilities. The video also touches upon the broader context of AI development, mentioning Elon Musk's departure from xAI and his new venture, as well as other AI-focused hedge funds. It references a Wall Street Journal article about Leo Aschenbrenner, a former OpenAI researcher who quickly amassed over $1.5 billion for his AI-focused hedge fund, "Situational Awareness." The video further mentions Google's AI updates, including new Imegen and Gemma models, and GPT-5's progress in playing games like Pokemon Red.

### AI Performance in IOI

- GPT-5 achieved gold medal-level performance
- ranked first among AI participants
- ranked sixth relative to humans
- achieved without specific IOI training
- demonstrated advanced general-purpose reasoning capabilities

### AI Industry News

- Elon Musk leaves xAI
- launches new venture
- Leo Aschenbrenner amasses $1.5 billion for AI hedge fund
- Google AI releases Imegen 4 and Gemma 3 models
- GPT-5 plays Pokemon Red

### AI Reasoning Capabilities

- Frontier models becoming more agentic
- taking on more tasks
- developing more powerful capabilities
- critical for studying and advancing AI safety

![Screenshot at 00:00: A split screen showing a video of robots running on a track and a man wearing headphones speaking to the camera.](https://ss.rapidrecap.app/screens/ikBODkpu7Jo/00-00-00.png)
![Screenshot at 00:27: A tweet from Elon Musk announcing Grok 4.20's potential to reach #1, mentioning xAI's workforce and aggressive release of new Grok versions.](https://ss.rapidrecap.app/screens/ikBODkpu7Jo/00-00-27.png)
![Screenshot at 01:21: A tweet from Igor Babuschkin discussing his last day at xAI and his vision for a new AI company with a different mission.](https://ss.rapidrecap.app/screens/ikBODkpu7Jo/00-01-21.png)
![Screenshot at 01:43: A screenshot of a "Humanity's Last Exam Benchmark Leaderboard" showing scores for Gemini 3.0, GPT-5, and Grok 4.](https://ss.rapidrecap.app/screens/ikBODkpu7Jo/00-01-43.png)
![Screenshot at 02:53: A tweet from Clad3815 showing GPT-5 playing Pokemon Red and achieving a new record, highlighting the efficiency improvement over previous models.](https://ss.rapidrecap.app/screens/ikBODkpu7Jo/00-02-53.png)
![Screenshot at 03:06: A "Milestones Chart" showing the progress of GPT-5 and GPT-3 vs. Pokemon Red, illustrating GPT-5's faster progression.](https://ss.rapidrecap.app/screens/ikBODkpu7Jo/00-03-06.png)
![Screenshot at 03:43: A Wall Street Journal article titled "Billions Flow to New Hedge Funds Focused on AI-Related Bets," featuring a 23-year-old former OpenAI researcher.](https://ss.rapidrecap.app/screens/ikBODkpu7Jo/00-03-43.png)
![Screenshot at 08:05: A tweet from Igor Babuschkin reflecting on his departure from xAI and his inspiration from Max Tegmark's question about building AI safely.](https://ss.rapidrecap.app/screens/ikBODkpu7Jo/00-08-05.png)
![Screenshot at 11:41: A tweet from Greg Brockman announcing gold medal-level performance for an OpenAI reasoning system in the International Olympiad on Informatics \(IOI\).](https://ss.rapidrecap.app/screens/ikBODkpu7Jo/00-11-41.png)
![Screenshot at 13:49: A screenshot of the IOI leaderboard showing the ranking of different AI models, including OpenAI's model at rank 5.](https://ss.rapidrecap.app/screens/ikBODkpu7Jo/00-13-49.png)