Post-Training and Deploying Open Source Reasoning Models in Foundry | DEM321
Fine-tuned small open-source reasoning models on Foundry can match frontier models for your domain at a fraction of the cost.
“This is almost like test-driven development for the age of agents.”
A Microsoft Build session demonstrates post-training open-source reasoning models in Azure AI Foundry, arguing that fine-tuned small models can hit a team's quality bar at a fraction of frontier-model cost as agent token consumption scales. The speakers push defining evals upfront as 'test-driven development for the age of agents,' enabling consistent quality measurement while swapping models. It matters because it reframes agent scaling around evals and cost-efficient open models rather than always reaching for the latest frontier model.