Microsoft’s new MAI Code 1.1 Flash gets crushed by Deepseek on both price and performance

What happened

Microsoft launched MAI Code 1.1 Flash, an updated code model integrated into GitHub Copilot. It claims to be 25 percent more token-efficient and operates at a quarter of the cost compared to the previous version. Despite these improvements, benchmark tests reveal that Deepseek’s V4 Flash model outperforms Microsoft’s offering on both price and performance metrics.

Why it matters

The new Microsoft model’s weaker performance relative to a cheaper competitor exposes a strategic tension. Microsoft promotes open AI tools broadly but continues to embed its own proprietary code models that lag behind open alternatives in efficiency and cost. This approach softens direct competition on price and quality in exchange for margin protection. For companies adopting GitHub Copilot or related Microsoft AI services, it means they may be paying more for a less capable model than alternatives available outside Microsoft’s ecosystem.

This dynamic pressures developers and businesses to reconsider lock-in risks with Microsoft’s AI offerings. Paying a premium that does not yield better performance increases operational costs and slows down development workflows. It also signals that proprietary AI models bundled into major platforms might be intentionally restrained, forcing customers to balance convenience against value.

What to watch next

Watch how Microsoft responds to Deepseek’s superior cost-efficiency in future updates. Will the company improve its AI models to close the gap, or lean further into proprietary control regardless of market competition? Also track how GitHub Copilot users and enterprise customers react—whether they demand better pricing and performance, or accept higher costs for integration benefits.

It is worth seeing if Deepseek or other open, competitive AI models gain traction by undercutting large incumbents. This could signal a shift where operators choose standalone AI tools over bundled, branded options.

AI Quick Briefs Editorial Desk