🧵1/9: 🚀 Introducing Llama3V, a new SOTA open-source VLM

We feature:
• Outperforms LLaVA 10–20%
• Training optimizations to train in under $500
• Comparable performance to GPT4-V, Gemini Ultra, Claude3

Check us out on:
• 🤗:
• Github:

As usual in long form at medium:

Built with: @maxaljadery, @siddrrshhuggingface.co/mustafaaljader…
github.com/mustafaaljader…
tinyurl.com/llama3v 🧵2/9: Llama3V combines state-of-the-art multimodal understanding with cost-effective training. Achieving a 10-20% boost over LLaVA, it delivers exceptional performance at a favorable price point.