The Model Question: Routing Models for Production Agents

​Demo night on model selection & routing for agents in production.

​Benchmarks are ever updating. The "best" model changes with every release week. So how do teams actually decide which model runs their agents in production?

​Five teams share live demos of how they evaluate, select, and route models in the real world, where quality, cost, latency, and reliability all pull in different directions.

​This room is for you if you're building agents or agent infra and asking "which model?"

​Program

​6:30 Doors open · food & drinks
7:00 GMI Cloud: routing every prompt to its best-fit model (featuring Yujing Qian from GMI Cloud)
7:15 How Notion picks the models behind Notion AI agents (featuring Hoda Noorian from Notion's AI team)
7:30 How Exa thinks about efficient model routing behind the scenes in our new product Agent (featuring João Adriano from Exa AI infra team lead)
7:45 Networking

​Who should come

​Developers and eng leads at AI-native startups building agents, agent infra, or anything that calls more than one model. Come with your hardest routing question.

​Hosted by

​GMI Cloud the inference platform for AI builders. GMI serves leading open and frontier models with high-throughput, low-latency infrastructure and is launching Model Navigator, a prompt-aware engine that recommends and routes to the best-fit model for every request.

​First 30 people get GMI Cloud swag.

Notion is where teams and agents think together. It brings your company’s knowledge, projects, meetings, and AI workflows into one collaborative workspace - so teams can find answers, automate busywork, and move work forward with shared context and controls. ⁠Startups get up to 3 months free of Notion Business + AI!