March 13, 2026
1M tokens, zero chill
1M context is now generally available for Opus 4.6 and Sonnet 4.6
No extra fees, mega memory: devs cheer while skeptics squint
TLDR: Anthropic turned on a 1‑million‑token memory for Opus and Sonnet at standard prices, plus higher media limits. Devs celebrate fewer cutoffs and simpler workflows, while skeptics demand proof it stays sharp near the max and note confusing access differences across plans—big news for anyone wrangling massive docs or code.
Anthropic just flipped the switch: its Opus 4.6 and Sonnet 4.6 now boast a 1 million-token memory at normal rates—no long-context surcharge—and can juggle up to 600 images or PDF pages. Translation for non-nerds: way bigger “memory” for your AI chats, without surprise fees. The crowd? Loud. The mood? Spicy.
Fans are calling it a workflow unlock for coding, legal docs, and long-running AI agents. One dev said multi-hour, multi-agent coding now fits in a single session. Another cheered that “standard pricing across the full 1M window” is the real deal—especially for Claude Code—if the AI stays coherent past 200k tokens. There’s battle chatter too: one commenter framed it as Anthropic’s move in the “agent wars” vs GPT, where keeping everything in one giant brain is a tactical flex.
But the skeptical chorus is strong. Veterans warn that “effective context”—what the AI actually uses well—can shrink near the limit. “Where’s the curve?” asks one pragmatist. Another cautions that those flashy “needle-in-a-haystack” tests don’t always mean better reasoning. Meanwhile, rollout weirdness sparked drama: one user suddenly had 1M context “(!!!)” without changing plans, another on Pro hit limits and cried foul. Cue memes about “AI with elephant memory,” “dumpster-load PDFs,” and a new sport called Prompt Hoarding. It’s equal parts victory parade and trust-but-verify vibes, and the comments are the main event.
Key Points
- •1M-token context windows for Claude Opus 4.6 and Sonnet 4.6 are now generally available on the Claude Platform at standard per-token pricing.
- •Pricing is set at $5 input/$25 output per million tokens for Opus 4.6 and $3 input/$15 output for Sonnet 4.6, with no long-context premium or multipliers.
- •Media limits are expanded to 600 images or PDF pages per request.
- •Claude Code (Max, Team, Enterprise) with Opus 4.6 now automatically uses the full 1M context, reducing the need for compaction.
- •At 1M tokens, Opus 4.6 scores 78.3% on MRCR v2 and Sonnet 4.6 scores 68.4% on GraphWalks BFS; customer examples report reduced compaction and improved agent performance.