Duration: 0:00

1,225 views

44

I rent a couple of GPUs from Hyperbolic Labs and try to run Llama 4 Scout as well as test out the new Gemini 2.5 Deep Research Model and their new text to video model: Veo 2.

Rent GPUs on Hyperbolic: http://hyperbolic.xyz/


Chapters:

1. Setting up Llama 4 Scout on Hyperbolic: (0:01-17:47)
• Kyle attempts to rent GPUs from Hyperbolic and install the Llama 4 Scout model, facing initial difficulties due to his lack of experience with these platforms.
• He uses ChatGPT and Claude to troubleshoot the installation process.

2. Testing Gemini 2.5 Deep Research: (17:47-57:00)
• Kyle switches to the Gemini 2.5 Deep Research model and gives it a prompt to perform a literature review on his PhD research topic.
• He's impressed by the speed and accuracy of the results, especially the retrieval of relevant information and specific values.

3. Veo 2 Text-to-Video Generation: (57:00-1:42:59)
• Kyle introduces Veo 2, Google's text-to-video model, and attempts to use it through the AI Studio collab workspace.
• He encounters issues with accessing the necessary APIs, requiring him to activate a free trial with Google Cloud.

4. Conclusion and Future Directions: (1:42:59-2:13:19)
• Kyle summarizes his experiences with the different models and tools, highlighting the potential of Gemini Deep Research for academic research.
• He discusses the limitations of current LLMs and mentions future possibilities, like improving non-circular motions in thin disk dynamical models.