【#Tech24H】Microsoft has released two coding models: MAI Code 2 and MAI Code 2 Flash. Microsoft states the models are tailored for routine enterprise programming tasks, with a strong focus on cost-performance: they deliver comparable output quality at less than half the price. Averaged across four widely adopted benchmarks, MAI Code 2 achieves 98% of GPT-6.1 Sol’s performance at just 40% of its cost, and reaches 96% of Claude Opus 5.5’s performance at only 5% of its cost. Furthermore, the models integrate natively into the Microsoft 365 ecosystem. In a key internal benchmark for Excel customer scenarios, Microsoft’s in-house model matches GPT-6 Luna in accuracy and token efficiency.

