CatalystTech/AI
openJiuwen X-Router debuts its self-evolving model routing technology, Ascend-friendly, Agents become more cost-efficient the more they run, with tests showing over 50% reduction in Token consumption.
openJiuwen X-Router debuts self-evolving routing with Ascend support and 50%+ Token reduction.
Context
What happened: openJiuwen released X-Router, a self-evolving model routing technology for AI Agents, built as part of the open source openJiuwen AI Agent platform. It adds a routing layer between Agents and models to pick a model per request.
Key numbers: PinchBench 147 tasks, 11 categories. Static routing scored 66.3% at $8.25 versus all Kimi-K2-Thinking at 71.36% and $11.11, a 25.7% cost cut.
Why it matters: The results indicate routing can approach full cloud model quality at much lower token cost, and the self-evolving design lets routing strategy adjust as models and workloads change.
What changed
openJiuwen released X-Router, a self-evolving model routing technology for AI Agents, built as part of the open source openJiuwen AI Agent platform. It adds a routing layer between Agents and models to pick a model per request.
Why it matters
The results indicate routing can approach full cloud model quality at much lower token cost, and the self-evolving design lets routing strategy adjust as models and workloads change.