skinny.

Latest / Neural Newscast

Microsoft MAI Models Beat Claude 4.6 in Blind Evals [Model Behavior]

Microsoft is making a significant move to control its own AI destiny with the debut of the MAI model suite at Build 2026. In this episode, we break down the performance claims of MAI-Thinking-1, which Microsoft says matches Claude Opus 4.6 on coding benchmarks. We contrast these internal figures with early hands-on reviews that characterize the models as unexceptional compared to established rivals. The segment also covers Google's new Managed Agents infrastructure, which allows for instant, headless Linux virtual machines to run model-generated code, and the privacy implications of OpenAI's…

Log in to save or queue this episode.

The skinny

The skinny isn't ready yet — notes appear once the transcript is processed.