Tokyo-based Rhymes AI has developed a new open-source multimodal language model (LLM) called Aria that can process text, code, images, and video. Aria stands out for its efficiency, employing a Mixture-of-Experts (MoE) framework that activates only the relevant experts for a specific task, reducing computational load. It combines the architectures of MoEs and multimodal LLMs, outperforming proprietary models and open-source heavyweights in benchmark tests. Aria's capabilities include analyzing financial reports, visualizing weather data, dissecting videos, and excelling in coding tasks. The model is released under the Apache 2.0 license, enabling developers to adapt and build upon it. Aria is a powerful addition to the open-source AI model space and represents a step toward achieving a fully open ChatGPT competitor.
- Content Editor ( decrypt.co )
- 2024-10-14
Meet Aria: The New Open Source Multimodal AI That's Rivaling Big Tech