claude-fable-5 — pricing, context window and limits
The previous Fable release, still served at Fable 5.1's per-token price but with cache reads at the standard rate — four times Fable 5.1's. It still accepts forced tool_choice, which Fable 5.1 rejects. Thinking is always on, and it requires 30-day data retention.
Price per million tokens
| Input | $10 |
|---|---|
| Output | $50 |
| Cache write, 5-minute TTL | $12.50 |
| Cache write, 1-hour TTL | $20 |
| Cache read | $1 (0.1x input) |
| Batch input / output | $5 / $25 |
Limits and behaviour
| Status | Current |
|---|---|
| Context window | 1M tokens |
| Max output | 128K tokens |
| Min cacheable prefix | 512 tokens |
| Effort levels | low, medium, high, xhigh, max |
| Effort default | high |
| Thinking | Always on — cannot be disabled |
| Forced tool_choice | Accepted |
| Platforms | Claude API |
| Released | 9 Jun 2026 |
Lifecycle
| API model name | claude-fable-5 |
|---|---|
| Retires no sooner than | 9 Jun 2027 |
| Successor | Claude Fable 5.1 |
The previous Fable, still served — and now the pricier way to cache
Fable 5 is no longer the top of the lineup: Fable 5.1 took its place in the same tier without changing the per-token input or output price. The difference that shows up on an invoice is the cache. A cache read on Fable 5 still costs the standard fraction of input, while Fable 5.1 reads at a fraction of that. For a long agentic session that re-reads a large cached prefix on every turn, that single line item is the whole reason to move — and the table above shows both rates side by side with the figures, not a paraphrase of them.
The one reason to stay a little longer
Fable 5 still accepts forced tool choice. Fable 5.1 rejects a tool_choice of any or tool
with a 400, so a harness that forces a specific tool call needs rewriting before it can move. See
forced tool_choice is not supported for the
replacement patterns. Everything else about the request surface carries over: thinking is always
on, and any explicit thinking configuration other than adaptive is rejected outright.
The retention requirement has not changed
Like its successor, Fable 5 requires a short data-retention window, which makes it unavailable to an organisation running under zero data retention. See Fable 5 is unavailable to zero-data-retention organisations for what that looks like and where to route instead.
Related
See migrating: Fable 5 to Fable 5.1 for the breaking changes in one place, and claude-mythos-5 for the same model behind Project Glasswing's gated access.
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).Could not confirm: Availability on Amazon Bedrock, Google Cloud and Microsoft Foundry is not stated in the source we used, so we do not claim it either way.
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.