"Auto Exacto" is now live, and on by default for tool-calling requests. Over the last few days, OpenRouter has reduced tool error rates by 15-90% across providers automatically. Here's how it works: https://t.co/BYGWkMkJNp
OpenRouter Launches Auto Exacto to Cut Tool-Calling Error Rates
ยท Updated
OpenRouter, an AI model routing platform, launched Auto Exacto, a routing layer that activates on every API request that includes tools. Instead of price-weighted provider selection, it reorders available providers based on throughput, tool-calling success rate, and internal benchmark data โ deprioritizing underperformers and promoting reliable ones. In its first few days, error rates dropped by 15โ90% across providers. OpenRouter plans to publish per-model accuracy and benchmark dashboards publicly.
The change matters most during new model launches โ exactly when provider variance spikes as inference engines finish patching chat templates and tool-call parsers. Previously, price-weighted defaults distributed tool calls across providers regardless of reliability. Auto Exacto makes the routing layer an active quality gate.
Developers building tool-calling AI agents benefit immediately โ no configuration required. To revert to price-weighted routing, set provider.sort to "price", use the :floor model variant, or update your account's default sort in preferences.
OpenRouter
@OpenRouter
29retweets
View on X