Hello, can someone help clarify? If the cost is 2× per million tokens, what’s the overall pricing difference between 4.6 and 4.6 Max? How should I think about this in the bigger picture?
Hey there,
Good timing on this question – the pricing for Max mode just changed a couple of days before your post.
Previously, Anthropic charged a 2x premium when input tokens exceeded 200k (the base context window), and Cursor passed that through. So the “2x” you’re referencing was real.
Recently, Anthropic removed the long-context premium entirely for Opus 4.6 and Sonnet 4.6. The full 1M context window is now available at standard pricing – no more 2x multiplier above 200k tokens. This has already been reflected in Cursor’s pricing.
Current pricing difference (Opus 4.6 vs Opus 4.6 Max):
-
Standard Opus 4.6 is billed at the model’s base API rates per token
-
Opus 4.6 Max applies a modest upcharge on those rates (see Max Mode for details)
-
Max mode also generates thinking tokens (billed at the output token rate), which can increase per-request cost further – depending on how much the model “thinks”
When Max mode is worth it:
-
You need the full 1M context window
-
You want extended thinking for complex multi-step reasoning
-
You’re using subagents with non-Composer models
If you’re doing simpler tasks that don’t need these features, standard mode is more cost-effective.
For full details: Models & Pricing | Max Mode