Anthropic releases Claude Opus 4.6

Anthropic released Claude Opus 4.6 yesterday, drastically expanding its context window to process millions of words in a single prompt. For UK SMEs, this update effectively eliminates the need to build expensive, complex retrieval-augmented generation (RAG) pipelines for internal search. Instead of paying developers to stitch together vector databases, teams can now just drop their entire company handbook and client history directly into the model.

Anthropic releases Claude Opus 4.6

Anthropic launched Claude Opus 4.6 this week, pushing the boundaries of how much data an AI model can hold in its short-term memory Anthropic. The update introduces a massive context window expansion.

Previous models required developers to chunk data, store it in a vector database, and retrieve only the most relevant snippets when a user asked a question. This method, known as RAG, was the standard way to build AI tools that could access private company data.

Claude Opus 4.6 changes the math. It can ingest vast amounts of text, code, and documents simultaneously without losing accuracy or forgetting details in the middle of the prompt. Anthropic claims the model achieves near-perfect recall across its entire context window Anthropic release notes. This means you can upload hundreds of PDFs, spreadsheets, and transcripts in one go, ask a highly specific question, and get a precise answer instantly.

The end of complex vector database projects

This fundamentally changes how a 50-person business should approach internal knowledge management. Until yesterday, building a custom AI tool to search your company data meant hiring engineers to set up RAG pipelines, manage vector embeddings, and fine-tune search algorithms. It was a costly project that often returned clunky results.

Now, you don't need that infrastructure. You can just feed your entire archive of HR policies, client contracts, or technical documentation straight into Claude Opus 4.6. The model reads it all fresh every time.

I think the industry is underestimating how quickly this will kill off mid-tier AI developer agencies. If you run a £10M manufacturing firm, you no longer need to buy a bespoke internal search tool. You just need a secure API connection to Anthropic and a basic chat interface.

The barrier to entry for building powerful internal tools has dropped from months of engineering to an afternoon of basic configuration. This shift heavily favours businesses that focus on organising their raw data over those trying to build clever search algorithms.

Three steps for your internal tools

1. Pause ongoing RAG development. If you're currently paying an agency to build a vector-based search tool for your internal documents, ask them to halt work. Evaluate if Claude Opus 4.6 can handle the workload natively.
2. Centralise your raw data. The new bottleneck isn't AI processing power, but data availability. Ensure your company documents, client notes, and process guides are stored in clean, readable text formats rather than locked inside proprietary software.
3. Test the context limits. Take a massive dataset, like your last five years of customer support tickets, and upload it directly into the Claude interface. Ask it to identify recurring complaints and see if the recall matches your manual reports.