Anthropic has officially launched Claude Opus 4.6, integrating its most sophisticated model into Microsoft Foundry to power the next generation of enterprise-grade AI agents. This update marks a significant shift toward autonomous work, where models do not just suggest code but manage the full development lifecycle from requirements gathering to long-term maintenance.

The model introduces a massive 1M token context window—essentially the model's short-term memory—allowing it to process entire codebases or hundreds of documents simultaneously. To help users manage these long conversations, a new feature called Context Compaction (beta) automatically summarizes older parts of a dialogue as token limits are reached, ensuring the AI maintains its focus without losing critical background information.

Perhaps most impressive is the advancement in computer use and adaptive thinking. Claude can now navigate software interfaces and fill out forms with higher accuracy, effectively acting as a digital employee. Meanwhile, adaptive thinking allows the model to dynamically decide how much reasoning effort to spend on a problem. It can perform simple tasks instantly or 'think' more deeply on complex logical challenges, optimizing both cost and performance.

By deploying on Microsoft’s infrastructure, companies like Adobe and Dentons gain access to these frontier capabilities while benefiting from enterprise-level security and governance. This integration bridges the gap between experimental AI and reliable production systems, enabling tools that can orchestrate dozens of sub-agents to solve high-stakes financial or legal tasks with minimal human oversight.

Anthropic has officially launched Claude Opus 4.6, integrating its most sophisticated model into Microsoft Foundry to power the next generation of enterprise-grade AI agents. This update marks a significant shift toward autonomous work, where models do not just suggest code but manage the full development lifecycle from requirements gathering to long-term maintenance.

The model introduces a massive 1M token context window—essentially the model's short-term memory—allowing it to process entire codebases or hundreds of documents simultaneously. To help users manage these long conversations, a new feature called Context Compaction (beta) automatically summarizes older parts of a dialogue as token limits are reached, ensuring the AI maintains its focus without losing critical background information.

Perhaps most impressive is the advancement in computer use and adaptive thinking. Claude can now navigate software interfaces and fill out forms with higher accuracy, effectively acting as a digital employee. Meanwhile, adaptive thinking allows the model to dynamically decide how much reasoning effort to spend on a problem. It can perform simple tasks instantly or 'think' more deeply on complex logical challenges, optimizing both cost and performance.

By deploying on Microsoft’s infrastructure, companies like Adobe and Dentons gain access to these frontier capabilities while benefiting from enterprise-level security and governance. This integration bridges the gap between experimental AI and reliable production systems, enabling tools that can orchestrate dozens of sub-agents to solve high-stakes financial or legal tasks with minimal human oversight.