Show HN: Echo – Fable-level results at 1/3 the cost using open-weight models
Summary
Echo pools open-weight models to decide per-request which models participate and how to combine outputs, aiming to beat single-model baselines while cutting inference cost to about one third. The author reports better aggregate results than the best individual model and similar to Fable, but notes deployment challenges due to needing post-hoc decision signals. The project includes a chat interface and API, with a public eval page and video explaining the approach, inviting readers to test and critique.