Anthropic Drops Long-Context Premium as Claude 4.6 Models Hit 1M Tokens

Anthropic just made a significant pricing move. The company announced Friday that its Claude Opus 4.6 and Sonnet 4.6 models now include the full 1 million token context window at standard rates—no premium pricing, no multipliers, no beta headers required.

A 900,000-token request now costs the same per-token rate as a 9,000-token one. That's $5/$25 per million tokens for Opus 4.6 and $3/$15 for Sonnet 4.6, unchanged from base pricing.

What Actually Changed

The practical implications matter more than the announcement itself. Developers previously needed to send beta headers for requests exceeding 200K tokens. That requirement is gone—existing code with beta headers will simply ignore them.

Media handling got a substantial upgrade too. Users can now process up to 600 images or PDF pages per request, up from 100. That's a 6x increase that opens up document-heavy workflows that were previously impractical.

The expanded context is live across Claude Platform, Microsoft Azure Foundry, and Google Cloud's Vertex AI.

Performance at Scale

Anthropic released benchmark data to back up the expansion. Opus 4.6 scores 78.3% on MRCR v2 at the full 1M token length—the highest among frontier models at that context size, according to the company. Sonnet 4.6 hits 68.4% on GraphWalks BFS under the same conditions.

These numbers matter because context length without recall accuracy is useless. Loading an entire codebase means nothing if the model can't find the relevant function buried 800K tokens deep.

Claude Code Gets the Upgrade

For Max, Team, and Enterprise users running Claude Code, Opus 4.6 sessions now automatically access the full 1M context window. The practical benefit: fewer conversation compactions and more of the session history preserved intact.

Previously, maintaining that context length required additional usage allocation. Now it's bundled in.

Why This Matters for Enterprise

The timing aligns with how enterprises are actually using these models. Opus 4.6, released February 5, 2026, has been deployed for security audits—Anthropic claims it discovered over 500 zero-day vulnerabilities in open-source code. Sonnet 4.6, which dropped February 17, handles the bulk of cost-sensitive workloads where speed matters more than maximum reasoning depth.

Eliminating the long-context premium removes a friction point for workflows that genuinely need massive input windows: full repository analysis, multi-year contract reviews, or agent traces that accumulate substantial context over extended sessions.

The move also pressures competitors. OpenAI and Google both offer extended context windows, but pricing structures vary. Anthropic betting that flat-rate simplicity wins developer mindshare isn't unreasonable.