Skip to main content

News Roundup 29

Created on September 11|Last edited on September 13
Open source foundation model Chai-1 outperforms AlphaFold 3
Chai-1 predicts molecular structures with advanced accuracy, enhancing drug discovery and disease research.
LongLLaVA scales multi-modal models to 1000 images efficiently
LongLLaVA processes up to 933 images on a single A100 80GB GPU, advancing multi-image AI tasks.
Roblox plans open source 3D foundational model
Roblox plans to launch a generative AI-powered 3D model, enhancing real-time interactive gaming experiences.


Mistral launches Pixtral 12B, a new multimodal AI model
Mistral unveils Pixtral 12B, a 12-billion parameter model for text and image tasks.
OpenAI releases o1-preview, a new model for complex reasoning
OpenAI's o1-preview aims to enhance problem-solving by spending more time on deep thinking.
Speculation on o1’s architecture
OpenAI’s o1 architecture might use multiple agents in Prover-Verifier Games to enhance reasoning.