Companies like Cognition and Cursor are proving a new pattern: using their proprietary user interaction data to fine-tune open-source models. This creates specialized AIs (e.g., for coding) that match or exceed general-purpose frontier models on specific tasks, while being significantly faster and cheaper to run.

The 'bigger is better' narrative is breaking down. For well-defined, structured tasks like coding and math, small models (e.g., 3 billion parameters) are now matching the performance of frontier models. This enables powerful, specialized AI to run on modest local hardware.

Relying solely on expensive frontier models is unsustainable. Vertical AI companies must build a portfolio of smaller, specialized models that match frontier performance on specific tasks but cost 100x less, effectively allocating intelligence where it's needed most.

Microsoft's research found that training smaller models on high-quality, synthetic, and carefully filtered data produces better results than training larger models on unfiltered web data. Data quality and curation, not just model size, are the new drivers of performance.

Instead of relying on expensive, omni-purpose frontier models, companies can achieve better performance and lower costs. By creating a Reinforcement Learning (RL) environment specific to their application (e.g., a code editor), they can train smaller, specialized open-source models to excel at a fraction of the cost.

Nadella describes a new frontier strategy: using a large, generalist model to generate initial traces for a specific task. These high-quality traces are then used to fine-tune a much smaller, specialized model, allowing it to achieve superior performance on that single task.

Microsoft's strategy lets companies customize proprietary models for specific tasks, achieving near-frontier performance at a fraction of the cost. This 'controlled tuning' approach is a powerful alternative to using expensive general models or relying on potentially inaccessible open-source options from abroad.

Microsoft is developing its own AI models from scratch, pitching them as cheaper and more effective for customized enterprise needs than leading models from its partner OpenAI or competitor Anthropic. This signals a strategy to control the full AI stack and compete directly on price.

For specialized, narrow tasks like classification, it's possible to distill the capabilities of a frontier model into a much smaller, fine-tuned model (e.g., under 1B parameters) and retain about 95% of the performance. This is a crucial strategy for managing cost and latency in production AI applications.