Anthropic officially announced the General Availability (GA) of the 1 million (1M) token context window for both Claude Opus 4.6 and Claude Sonnet 4.6.
The most significant part of the announcement is the removal of the “long-context premium.” Anthropic is now applying standard pricing across the entire 1M window, a move that undercuts major rivals who typically charge higher multipliers for high-context prompts.
The New Pricing Structure
A 900,000-token request is now billed at the exact same per-token rate as a 9,000-token one.
| Model | Input Price (per 1M tokens) | Output Price (per 1M tokens) | Status |
| Claude Sonnet 4.6 | $3 | $15 | Now standard 1M context. |
| Claude Opus 4.6 | $15 | $75 | Now standard 1M context. |
- No Beta Header Required: Developers no longer need to use specific beta headers to access the 1M window.
- Multiplier Removal: Third-party tools like Cursor have already updated their pricing to reflect this, removing the “2x multiplier” previously applied to inputs exceeding 200,000 tokens.
Key Technical Upgrades: “Context Compaction”
To solve the industry-wide problem of “context rot” (where models lose track of instructions as windows fill), Anthropic introduced Context Compaction.
- Automated Summarization: When a conversation approaches the 1M limit, the API automatically summarizes the oldest parts of the chat into a compressed “state” to preserve signal.
- Retrieval Performance: On the MRCR v2 (Multi-needle Retrieval) benchmark, Opus 4.6 achieved 76% accuracy at 1 million tokens, a massive leap from the ~18% seen in previous generations.
- Media Limits: The update also expands media processing capabilities to 600 images or PDF pages in a single prompt.
The Competitive Landscape (March 2026)
Anthropic’s pricing move is a direct challenge to the tiered pricing models used by its main competitors:
- Google Gemini 3.1 Pro: Currently applies a price hike for prompts exceeding 200,000 tokens.
- OpenAI GPT-5.4: Features a pricing increase once inputs cross the 272,000-token mark.
Availability
- Claude.ai: 1M context is now included by default in Claude Code for Max, Team, and Enterprise users.
- API & Cloud Platforms: The update is live on the Claude Platform, Amazon Bedrock, and Google Cloud Vertex AI.
