Echo: Fable-Level Performance with Open-Weight Models
Boost AI performance while cutting costs by two-thirds.
1What is Echo?
Echo is a project that combines multiple open-weight AI models into a single system. It aims to achieve higher performance by selecting the most suitable models for each task and merging their outputs.
2How Does It Work?
Echo automatically determines how much computational power to allocate, which models to use, and how to combine their outputs for each request. This allows for better results with fewer resources in some tasks and collaboration of multiple models in others.
3Model Diversity
Echo discovered the complementary features of different models. A model that appears weaker overall can be quite useful in specific tasks or combinations.
4Performance Comparison
In initial evaluations, Echo outperformed the best individual model in its pool. It also achieved results close to a strong comparison system like Fable while reducing costs by two-thirds.
5Current Challenges
Echo can sometimes make incorrect allocation or combination decisions. The developer is working to understand these errors and test whether the same approach holds for coding and agentic tasks.
6Try and Provide Feedback
Echo offers a chat interface and an OpenAI-compatible API. Users are encouraged to test the system and report any issues they encounter. This feedback will help further improve the system.