claude-opus-4-8 — pricing, context window and limits
Same price as Opus 5 and the same request surface as 4.7. Thinking is OFF unless you ask for it: omitting the thinking field runs without thinking, the opposite of every Opus after it. It is also a documented fallback target for Fable 5.1 and Opus 5.5 refusals.
Price per million tokens
| Input | $5 |
|---|---|
| Output | $25 |
| Cache write, 5-minute TTL | $6.25 |
| Cache write, 1-hour TTL | $10 |
| Cache read | $0.50 (0.1x input) |
| Batch input / output | $2.50 / $12.50 |
| Fast mode input / output | $10 / $50 |
Limits and behaviour
| Status | Current |
|---|---|
| Context window | 1M tokens |
| Max output | 128K tokens |
| Min cacheable prefix | 1,024 tokens |
| Effort levels | low, medium, high, xhigh, max |
| Effort default | high |
| Thinking | Off unless requested (adaptive) |
| Forced tool_choice | Accepted |
| Platforms | Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry |
Lifecycle
| API model name | claude-opus-4-8 |
|---|---|
| Retires no sooner than | 28 May 2027 |
| Successor | Claude Opus 5.5 |
The fallback target, not just another point release
Opus 4.8 shares Opus 5's per-token price and Opus 4.7's request surface, with one behavioural
difference that matters more than it looks: thinking is off unless you ask for it. Omit the
thinking field and the request runs without thinking — the opposite of every Opus released since.
It is also one of the documented fallback targets when Fable 5.1 or Opus 5.5 declines a request,
so a fallback chain can route traffic here without anyone choosing it deliberately.
Why the thinking default matters most on a fallback
Code tuned against Opus 5.5, where thinking is always on, can be surprised the one time a fallback lands here and nothing thinks unless the request says so. A reasoning-heavy request that arrives on this model through a fallback should set adaptive thinking explicitly rather than rely on the destination model's default. The reverse also holds: a fallback model cannot read the declining model's thinking blocks, so the first turn after a switch re-plans without them.
Fast mode is priced now
This model supports fast mode on the Claude API at its own premium rate, which Anthropic now publishes — the table above renders it from the same dated source as the standard rates. See how fast mode changes the price of Opus for when that premium earns its keep.
Related
See migrating: Opus 4.8 to Opus 5.5 for the move the migration guide now recommends, and migrating: Opus 4.8 to Opus 5 for the intermediate step.
Verified 2026-09-30 against claude-api skill — shared/models.md, model-migration.md + SKILL.md model table and Anthropic's pricing and model-deprecation pages (canonical: https://platform.claude.com/docs/en/about-claude/models/overview, https://platform.claude.com/docs/en/about-claude/pricing, https://platform.claude.com/docs/en/about-claude/model-deprecations).
Verified 2026-10-28 against Anthropic's model-deprecations status table and pricing page (canonical: https://platform.claude.com/docs/en/about-claude/model-deprecations, https://platform.claude.com/docs/en/about-claude/pricing).Could not confirm: These dates apply to the Claude API, Claude Platform on AWS, Claude Models on Google Cloud (Anthropic managed) and Microsoft Foundry. Amazon Bedrock and Claude Models on Google Cloud (Google managed) keep their own retirement schedules, which we do not track.