Anthropic Launches Claude Opus 4.6: Agent Teams and 1M Token Context

On February 5, 2026, Anthropic introduced Claude Opus 4.6. This major update to the flagship model shifts the focus from simple conversational interaction to executing long-running, autonomous "agentic workflows."

Technical Advancements

The model incorporates several architectural improvements designed for large-scale tasks:

• 1M Token Context (Beta): For the first time in the Opus class, support for a context window of up to 1 million tokens has been announced. This feature is currently in beta and is designed for scenarios requiring the ingestion of entire codebases or massive document archives without losing fidelity.
• Adaptive Thinking: A new "adaptive thinking" capability allows the model to autonomously determine when a task requires deep analysis and extended "thinking time" versus a quick response. Developers can manage this balance via a configurable effort parameter.
• Context Compaction: To maintain performance during extended sessions, the model can now summarize its own history ("context compaction"), freeing up space for new information while retaining essential context.

Performance and Benchmarks

Anthropic states that Opus 4.6 demonstrates state-of-the-art results on benchmarks evaluating "economically valuable knowledge work," including complex tasks in finance and law. According to company tests (such as GDPval-AA), the new model outperforms other current frontier models in these professional domains.

Tools for Developers and Enterprise

The update integrates the AI more deeply into professional environments:

• Claude Code: Within Anthropic’s specialized developer tool, users can now deploy agent teams. This allows multiple AI agents to work in parallel on a single objective—for example, one agent can write code while another simultaneously prepares tests or documentation.
• Office & Document Workflows: The model’s capabilities with corporate software have been significantly enhanced. Claude is now more effective at analyzing unstructured data within spreadsheets for financial modeling. Additionally, it features improved abilities to generate presentation materials, creating slide structures and content that align closely with professional standards and templates.