ClaudeHowSupport Us

claude-opus-4-6 — pricing, context window and limits

Statuscurrent
Input / MTok$5
Output / MTok$25
Context window1,000K tokens
Max output128K tokens
Min cacheable prefix4,096 tokens
Effort levelslow, medium, high, max
Thinkingadaptive-opt-in
PlatformsClaude API, Amazon Bedrock, Google Cloud, Microsoft Foundry

The last of the old request shape

Opus 4.6 is the final Opus generation that still accepts the older sampling controls (temperature, top_p, top_k) and the deprecated thinking-budget mechanism, and it's the last one without the highest effort tier available on every later generation. Every Opus released after it rejects all four of those older fields outright, which makes this model the actual boundary line for a whole category of migration work, not just an incremental step along it.

The highest cache minimum in the current lineup

This model's minimum cacheable prefix sits at the top of the range across every current and recent model — a prompt engineered to cache reliably here will cache reliably almost everywhere else in the lineup too, but the reverse isn't guaranteed. If you're migrating away from this model, expect some prompts to newly qualify for caching on the destination model that never cached here at all.

What breaks first on a migration away from this model

The older sampling fields are usually what surfaces first during testing, since removing them produces an immediate, hard error rather than a subtle behaviour change. See temperature, top_p and top_k are not supported on this model and budget_tokens is not supported on Opus 4.7 and later for both fields this model is the last to accept.

See migrating: Opus 4.6 to Opus 5 for the full diff, and what actually changed from Opus 4.6 to Opus 5 for the narrative version of the same migration.

Verified 2026-08-08 against claude-api skill — shared/models.md + SKILL.md model table (canonical: https://platform.claude.com/docs/en/about-claude/models/overview, https://platform.claude.com/docs/en/pricing).