Models9 Feb 2026

Opus 4.6 and the million-token context: feed it the whole business

This week's release reads roughly 700,000 words at once. That changes the questions you can ask.

Anthropic released Claude Opus 4.6 on February 5, with a context window of up to a million tokens in beta — roughly 700k words. For a small business, that's not a spec, it's a new capability: whole-picture questions.

  • Drop in a year of support emails: "What are the ten most common complaints, ranked, with example quotes?"
  • All your contracts at once: "Which have non-standard payment terms or auto-renewals in the next 90 days?"
  • Your entire website: "Find every inconsistency in pricing, dates and product names."

Two cautions: long context costs real money per question (batch your questions — ask five in one pass), and "it fits" doesn't mean "it's organized" — a model reasons better over clean, labeled documents than over a junk drawer. Same as people.

← Previous Let AI clean the customer list you've been avoiding MCP, explained without the acronyms Next →

Want this done for you?

I set up AI agents for small and medium businesses — start with a free automation audit.

Message me on Telegram