Open-source small reasoning models heat up

Over the past seven days, open-source communities concentrated on small reasoning models. Several 7B to 20B parameter models claim to approach the capability of larger models in coding, function calling, and local deployment.

The common theme is fitting reasoning and tool use into a smaller context budget, making it practical to run multi-step agents on consumer hardware. For individual developers, the bar for running a capable local model is clearly dropping.

Source: example link; replace with real Tavily sources when generating.