HN: Anthropic Makes 1M Context Standard for Opus 4.6 and Sonnet 4.6

Anthropic’s March 13 announcement landed hard on Hacker News, where the post had 1,110 points and 477 comments at crawl time. The headline was simple: Opus 4.6 and Sonnet 4.6 now support a 1M-token context window in general availability. What made the story high-signal was not only the size of the window, but the way Anthropic changed the economics around it.

According to Anthropic, standard pricing now applies across the entire 1M window instead of requiring a separate long-context premium. The company lists Opus 4.6 at $5 input and $25 output per million tokens, and Sonnet 4.6 at $3 input and $15 output per million tokens. Anthropic also expanded media limits to 600 images or PDF pages, up from 100, and said the feature is available on Claude Platform as well as through Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. For Claude Code, Max, Team, and Enterprise users on Opus 4.6 now default to the 1M window automatically.

Why HN treated this as more than a model spec

• The pricing change matters as much as the context number because it turns long context from a premium exception into a normal operating mode.
• The larger media limit makes the feature more useful for code review, large PDFs, discovery workflows, and multi-document analysis.
• Automatic availability in Claude Code reduces the need for aggressive compaction before a workflow even begins.

The practical impact is not that every prompt should suddenly be inflated to 1M tokens. It is that teams working with large repositories, research corpora, legal records, and knowledge-heavy internal docs can keep far more source material in one pass without paying a separate multiplier. That changes the design space for agentic tools. Instead of building retrieval, chunking, and summarization layers first, developers can test a more direct workflow and add compression only where it still proves necessary.

That is why the HN response is worth preserving as a community article. The discussion was attached to a concrete product change with immediate consequences for how people build coding agents, research assistants, and document-analysis systems. Anthropic is effectively arguing that long context is no longer a niche benchmark feature. Between standard pricing, broader platform support, and the default behavior inside Claude Code, the company is trying to make 1M context feel like infrastructure.

Source: Anthropic · Community discussion: Hacker News