Weave adds model routing and self-hosting

This round is pushing Weave from a dashboard product into core AI infrastructure for engineering teams. The new spend goes into two hard things that matter in enterprise buying, measuring real engineering output instead of raw activity, and deciding which model should handle each coding task based on cost and result quality. Self hosted deployment is the other unlock, because large companies will not pipe source code and editor data into a shared SaaS by default.1 2 3 4

• Weave already connects to GitHub, Jira, and CI/CD systems, then scores pull requests for complexity, work type, review quality, and AI generated code usage. Better engineering output systems means improving that scoring layer so teams can compare human work, AI assisted work, and delivery speed in one place, which is what budget owners need before expanding Copilot or model spend.1 2
• Model routing is becoming a second product, not just an internal feature. Weave describes routing as choosing the best model for each request using downstream performance data, and its router repo exposes local and self hosted operation plus cost analytics. That turns Weave from an observer of AI coding into a traffic controller for where token dollars go.2 3 5
• Self hosted environments are mainly about regulated buyers and sensitive codebases. Comparable developer and AI tools use self hosted or air gapped editions to win defense, healthcare, financial services, and other security constrained accounts. For Weave, that expands the market from startup engineering teams into enterprises that need code and model traffic to stay inside their own network.1 2 6 7

The next step is a tighter bundle where Weave both measures AI coding ROI and actively improves it. If that bundle works, the company can sell not just analytics seats at $50 per engineer per month, but a higher value enterprise system that governs code data, model choice, security, and AI spend in one deployment.1 2 3 4 5