Claude Opus 4.6 adds 1M context, agent teams

This is less a model bump than Anthropic turning Claude into a multi-agent work stack.

• –Benchmarks across coding, agentic search, and knowledge work suggest Opus 4.6 is a real capability jump, not just a bigger window.
• –The 1M-token context window matters most when paired with compaction and 128k outputs, because that keeps long sessions usable instead of brittle.
• –Agent teams are the strategic shift: Claude Code moves from one assistant at a time to parallel work that can split tasks, specialize, and merge results.
• –Anthropic is tying the model to real work surfaces like Excel, PowerPoint, and Claude Code, which signals a broader productivity play rather than a narrow coding launch.
• –The practical tradeoff is cost and control, so effort tuning and strong evals will matter as much as raw model quality.