Microsoft MAI Code 1.1 Flash Falls Behind Competitors on Price and Performance

Microsoft has released its new MAI Code 1.1 Flash model for GitHub Copilot, but it falls short on both price and performance compared to competitors.

The model claims to write better code, is 25% more token-efficient, and costs a quarter of its June predecessor, with developers accepting 4% more of its output after training on 'hundreds of thousands of reinforcement-learning environments in GitHub Copilot.'

However, benchmarks show that MAI Code 1.1 Flash trails Deepseek-V4-Flash in both price and performance, despite being cheaper than before.

This contradicts Microsoft's recent push to paint itself as an open AI champion, with the company instead choosing to invest in a weaker, pricier in-house model that is proprietary and unlikely to be released under open weights.