A new approach to improve search relevance in AI by fine-tuning models to better capture and rank context for accurate answers.

Meta released powerful multilingual foundation models (8B, 70B, 405B) with integrated image, video, and speech capabilities using a compositional approach.

A framework combining multiple LLMs, achieving state-of-the-art results by leveraging collaborative expertise from different models.

The OpenAI "Strawberry" widely known as 'o1 model' is specifically designed to handle more complex reasoning and planning tasks, it is achieved through an expanded architecture and enhanced attention mechanisms for improved context retention over longer interactions. Reinforcement learning is used to fine-tune responses, making them more aligned with user preferences and enhancing conversational quality. This (RL) enables Strawberry to minimize hallucinations and produce more reliable results. It is still in its early stages and limited to certain user groups