OpenAI announced that GPT-5.6 is now the preferred model across Microsoft 365 Copilot—Word, Excel, PowerPoint, Copilot Chat, and Cowork. The rollout includes three variants: Sol (flagship), Terra (balanced for everyday work), and Luna (most cost-efficient). Sam Altman reported to CNBC that GPT-5.6 delivers 54% better token efficiency for agentic coding tasks, a massive leap that alters the economics of AI coding-agent deployments.
GPT-5.6 Sol achieves state-of-the-art results across coding, knowledge work, cybersecurity, and science while outperforming prior and competing frontier models with fewer tokens and lower estimated cost. On OSWorld 2.0 (agentic task evaluation), it surpasses Claude Fable 5 while using 85% fewer output tokens. On BrowseComp (web browsing agents), GPT-5.6 Sol reaches 92.2%, a new high. In app-building conversations, GPT-5.6 used 22% fewer input tokens and 23% fewer output tokens than GPT-5.5 while staying competitive on multi-turn work.
For Microsoft 365 customers, the deployment is seamless: no new interface, no data exports. Copilot automatically selects the right model variant based on prompt complexity. For complex or open-ended prompts, Copilot routes to deeper reasoning; for routine questions, it uses faster throughput. The timing signals pressure on OpenAI to reclaim performance ground after Anthropic and Google have emphasized coding and efficiency.
For architects and AI platform teams: Token efficiency is now a business metric, not an implementation detail. A 54% reduction in token spend per coding task changes unit economics for agent-heavy workflows at scale. If you're building or deploying multi-agent systems for knowledge work, now is the time to re-baseline token burn against GPT-5.6 and benchmark your stack accordingly. Microsoft's locked-in integration means Copilot will be the default for millions of Office users; competitors need to demonstrate equivalent or lower cost-per-task to justify switching.