Show HN: Echo – Fable-level results at 1/3 the cost using open-weight models

	Show HN: Echo – Fable-level results at 1/3 the cost using open-weight models
	451 points by adam_rida 1 day ago \| 214 comments
	tl;dr: Echo is a routing system that dynamically allocates inference across a pool of open-weight models (GLM-5.2, Kimi K2.7, etc.), deciding per-request which models participate and how their outputs are combined. The author claims it outperforms any single model in its pool and matches Fable's aggregate results at roughly one-third the inference cost. A chat UI and OpenAI-compatible API are available for testing, with ongoing work on failure analysis and extending to coding/agentic tasks.
	HN Discussion: ↓The landing page uses dark patterns and requires signup before demonstrating any value, driving users away ↓The launch is premature, lacking benchmarks, model transparency, and basic trust signals like a no-credit-card trial ↑Multi-model routing/orchestration is a promising future architecture that could replace the pursuit of a single best model ↓The cost savings claim is undermined by subsidized flat-rate plans and by cache-breaking inefficiencies of routing across models ~The routing approach adds complexity around auditability, context memory, and debugging LLM behavior