So you want to use OpenRouter?
Summary
The article shares practical, experience-based insights on using OpenRouter with open-source LLMs, drawn from the author's Olly AI assistant implementation. It highlights real-world pitfalls such as provider variance in benchmarks, inconsistencies across vision models, optional reasoning effort, quantization effects, and tool-calling reliability, offering concrete guidance and references for practitioners. The content is valuable for technical readers interested in evaluating AI tooling, integration patterns, and reliability considerations when routing models across providers.