Claude Code Gets 1M Context Window by Default for Paid Plans

Q: What Changes with General Availability

The update brings several key improvements: One price, full context window — No long-context premium. Your standard rate applies whether you use 10K or 900K tokens. Full rate limits at every context length — Standard account throughput applies across the entire window, so you are not throttled for using more context. 6x more media per request — Up to 600 images or PDF pages per request, up from the previous limit of 100. No beta header required — Requests over 200K tokens now work automatically. If you are already sending the beta header, it is simply ignored.

Q: How Does Retrieval Accuracy Hold Up?

Anthropic claims that Opus 4.6 scores 78.3% on MRCR v2 (Multi-needle Retrieval in Context Reasoning), which they say is the highest among frontier models at that context length. While that is impressive on paper, it is worth noting that this still means roughly 1 in 5 retrieval attempts may not find the right information in a full 1M token context. That said, the practical impact depends heavily on the use case. For code review, where the relevant files are typically clustered together, the accuracy is likely much higher than the benchmark suggests. For scattered cross-referencing across a massive context, the limitations may be more noticeable.

By Jaspal March 14, 2026 Updated: March 14, 2026

Anthropic Makes 1M Context Default for Claude Code Users

Anthropic has announced that the full 1 million token context window is now generally available for Claude Opus 4.6 and Sonnet 4.6, with Claude Code users on Max, Team, and Enterprise plans getting it as the default setting. Previously, accessing the 1M context window required additional usage costs.

The move effectively removes the long-context premium that Anthropic had been charging. Standard pricing now applies across the entire context window — $5/$25 per million tokens for Opus 4.6 and $3/$15 for Sonnet 4.6. A 900K-token request is billed at the same per-token rate as a 9K one.

What Changes with General Availability

The update brings several key improvements:

One price, full context window — No long-context premium. Your standard rate applies whether you use 10K or 900K tokens.
Full rate limits at every context length — Standard account throughput applies across the entire window, so you are not throttled for using more context.
6x more media per request — Up to 600 images or PDF pages per request, up from the previous limit of 100.
No beta header required — Requests over 200K tokens now work automatically. If you are already sending the beta header, it is simply ignored.

Why This Matters for Claude Code Users

For developers using Claude Code, this is a significant quality-of-life improvement. Opus 4.6 sessions can now use the full 1M context window automatically, which means fewer context compactions and more of the conversation kept intact during long coding sessions.

Context compaction has been a persistent pain point for developers working on large codebases. When the context window fills up, the system has to summarize and compress earlier parts of the conversation, which can lead to lost details and the AI “forgetting” important context. With 1M tokens available by default, these compaction events happen far less frequently.

How Does Retrieval Accuracy Hold Up?

Anthropic claims that Opus 4.6 scores 78.3% on MRCR v2 (Multi-needle Retrieval in Context Reasoning), which they say is the highest among frontier models at that context length. While that is impressive on paper, it is worth noting that this still means roughly 1 in 5 retrieval attempts may not find the right information in a full 1M token context.

That said, the practical impact depends heavily on the use case. For code review, where the relevant files are typically clustered together, the accuracy is likely much higher than the benchmark suggests. For scattered cross-referencing across a massive context, the limitations may be more noticeable.

Platform Availability

The 1M context window is available today on the Claude Platform natively, as well as through Amazon Bedrock, Google Cloud’s Vertex AI, and Microsoft Foundry. This broad availability across cloud providers means teams already using these platforms can take advantage of the expanded context without migration.

The Bottom Line

Anthropic removing the long-context premium and making 1M tokens the default for paid Claude Code users is a welcome move, even if it raises the question of why it was an upsell in the first place. For developers spending hours in Claude Code sessions, fewer compaction events and a larger working memory should translate to genuinely better coding assistance. Whether the 78% retrieval accuracy at scale is good enough for production-critical work remains to be seen, but for most day-to-day development tasks, this is a meaningful upgrade.