What happened
Workweave released an AI model router on GitHub, designed for agentic systems, routing prompts to optimal models from providers like Anthropic, OpenAI, Gemini, and OpenRouter in under 50 milliseconds. Built by Weave, it claims 40-70% cost reduction through dynamic model selection via an Avengers-Pro 1 cluster scorer. The router supports various APIs, BYOK encryption, and OTLP observability, integrating with tools such as Claude Code, Codex, and Cursor via a single endpoint change.
Why it matters
This model router fundamentally shifts the unit economics of AI inference for platform engineers and CTOs. By dynamically selecting the most cost-effective model per request, it reduces API spend and simplifies multi-provider deployments, offering a direct mechanism for cost reduction. Security architects benefit from the Bring Your Own Key (BYOK) encryption, maintaining control over sensitive provider credentials. This development follows recent efforts like OpenRouter Fusion, which also aims to halve AI costs, indicating a growing industry focus on optimising model consumption.
