NewsBite analysis · AI, Cloud, Programming, Startups
The Decoder
Microsoft's new MAI Code 1.1 Flash gets crushed by Deepseek on both price and performance
Microsoft has launched MAI Code 1.1 Flash, a code model for GitHub Copilot, but it faces significant challenges against competitors, particularly Deepseek. While claiming to be more efficient and cheaper, the new model does not perform as well in benchmarks and is priced higher than Deepseek-V4-Flash-0731. Despite offering some improvements over its predecessor, the model's capabilities appear limited, raising questions about Microsoft's strategy of favoring in-house models over more effective alternatives. This could impact user choices and market dynamics as Microsoft potentially prioritizes its proprietary offerings.
MonitorConfidence: MediumShort term horizonReputational risk: High
Published 12 Aug 2026, 12:10 UTC2 min read
Visit original article ↗Lead image — The Decoder
Summary
TL;DR
- MAI Code 1.1 Flash claims to be 25% more token-efficient than its predecessor.
- Deepseek outperforms Microsoft's new model on both price and performance.
- Microsoft buried detailed benchmarks, focusing on vague improvement metrics.
- The new model's launch reflects a shift in Microsoft's strategy for AI development.
- Developers accepted 4% more of the outputs from the MAI model.
Interesting facts
- Microsoft's internal strategy appears focused on cost-cutting rather than performance.
- The new model may become the default in Microsoft's ecosystem, reducing user choice.