Anthropic just made its full 1 million token context window generally available for Claude Opus 4.6 and Sonnet 4.6 — at standard pricing, with no long-context premium. A 900K-token request now costs the same per-token rate as a 9K one.
This is a meaningful shift. Until now, working with very large contexts meant either paying a multiplier or using beta headers. Now the full window is the default: no special flags, no extra cost, and rate limits apply uniformly across the entire range. Media limits also jumped from 100 to 600 images or PDF pages per request.
Why it matters:
- Developers and teams can feed entire codebases, full legal case files, or long agent traces into a single session without chunking, summarizing, or losing context.
- Claude Code (Anthropic's CLI agent) benefits directly — fewer compaction events mean longer, more coherent coding sessions.
- Opus 4.6 leads frontier models on the MRCR v2 long-context retrieval benchmark at 78.3%.
- Available on Claude Platform, Microsoft Azure Foundry, and Google Cloud Vertex AI.
Also in the news: Meta delayed its next major AI model (codenamed "Avocado") from March to at least May, reportedly because performance falls short of rivals like Google. And Ben Affleck's AI startup was acquired by Netflix for around $600 million.
Relevant Links
- Anthropic blog post: https://claude.com/blog/1m-context-ga
- Hacker News discussion: https://news.ycombinator.com/item?id=47367129
- The Verge AI roundup: https://www.theverge.com/ai-artificial-intelligence