Over the past weekend we partnered with Meta and Cerebral Valley to host a Lllama 4 specific hackathon.
Over 300 developers/hackers attended.
5 things we learnt/saw/experienced:
-
We saw many cool projects especially those around Llama 4’s multimodal capabilities. Video processing, Virtual AI Agents and Mixture of Experts were all topics that were very popular.
-
The developer community was impressed with how fast SambaNova can run inference on Llama 4 Maverick. We showcased this during our workshop on stage.
-
Shack15 is a beautiful place to host a hackathon. If you haven’t been there plan on visiting!
-
With the help of open-source tools, the processes of training, fine-tuning, and distillation are becoming increasingly streamlined. Additionally LLM Evals, are becoming more standardized and accessible, enabling more efficient assessment and improvement of these models
-
With Llama4 + Vibe coding it’s now easier than ever to build amazing projects within 24 hours. Judging hackathon projects is even HARDER now! The final judging took over an hour of intense deliberation to decide the winner.
We are looking forward to partnering with more hackathons and meeting you at your local community. Let us know if you know of a hackathon that we should attend/partner/sponsor in the future!