RC RANDOM CHAOS

MAI Code 2 Delivers Frontier Coding on a Budget

· via Hacker News

Original source

MAI Code 2 and Code 2 Flash

Hacker News →

MAI offers two code generation models designed for different development needs. MAI Code 2 handles complex enterprise tasks including production deployment, multi-file implementations, repository-wide analysis, debugging, and UI work. MAI Code 2 Flash targets high-volume routine work such as code explanations, targeted edits, and quick fixes. The company notes these models are not currently optimized for fundamental research or long-horizon agentic work.

Benchmark testing across four enterprise coding tasks shows MAI Code 2 achieves 98% of GPT-6 Sol’s performance at 40% of the cost and 96% of Claude Opus 5.5’s performance at 5% of the cost. In blind frontend application evaluations, reviewers preferred MAI Code 2 over GPT-6 Sol 64% of the time. MAI Code 2 Flash delivers comparable performance to GPT-6 Luna and Claude Haiku 5.5 at 65% and 20% of their respective costs.

MAI Code 2 is becoming a core component of Microsoft’s product suite, with an optimized variant (MAI Work Flash) powering M365 and Fabric IQ. Performance on Excel customer scenarios matches GPT-6 Luna’s accuracy and token efficiency. Both models are available through GitHub Copilot, with Code 2 Flash in limited preview on Microsoft Foundry, and upcoming availability on Open Router, Vercel, Hermes, Open Code, and Kilo.

Read the full article

Continue reading at Hacker News →

This is an AI-generated summary. Read the original for the full story.