The model release

Microsoft AI, the company's AI research division, made the 2.8 trillion-parameter model public, drawing attention to its scale. Parameters — the internal values that determine how a model processes information — serve as a rough measure of a model's capacity, and a parameter count of 2.8 trillion places the model among the largest ever publicly disclosed.

Efficiency as the headline claim

Alongside the size of the model, the announcement highlighted a dramatic improvement in efficiency. Microsoft AI presented the efficiency gains as a central selling point of the release, indicating that the company is aiming to show that very large models can be made to run more economically rather than simply growing bigger.

What remains unclear

The available information does not specify which benchmarks the efficiency figures are based on, how the model will be made available to users or developers, or what specific architecture choices enable the gains. Those details remain unconfirmed in the disclosure reviewed so far.

The claim that Microsoft AI has released a 2.8 trillion-parameter model with substantially improved efficiency stands as True.