Claude Fable 5
Anthropic's most capable widely released model — launched 9 June 2026, pulled three days later under US export controls, and back globally since 1 July. No other 2026 model has a stranger release story.
Claude Fable 5 specs at a glance
| Maker | Anthropic |
| First released | 9 June 2026 |
| Globally available again | 1 July 2026 (after export controls lifted 30 June) |
| Context window | 1,000,000 tokens (default and maximum) |
| Max output | Up to 128,000 tokens |
| API pricing | $10 / $50 per 1M input / output tokens |
| Safeguards | Falls back to Opus 4.8 on some topics (<5% of sessions) |
| Platforms | Claude API, claude.ai, Claude Code, AWS Bedrock, Google Cloud, Microsoft Foundry |
Calculate API costs for this model → Compare its monthly bill against 15 other rate cards.
The strangest release timeline of 2026
Fable 5 became available on 9 June 2026 across the Claude API, Amazon Bedrock, Google Cloud and Microsoft Foundry. Three days later, the US Department of Commerce instructed Anthropic to pull the model globally over national-security concerns tied to export controls — an unprecedented intervention against a generally available frontier model.
On 30 June 2026 the export controls were lifted, and from 1 July Fable 5 returned to users globally on the Claude Platform, claude.ai, Claude Code and Claude Cowork, with Amazon Bedrock restoring access under strengthened guardrails. Anthropic has also opened a bug bounty for Fable 5 jailbreaks on HackerOne.
Capability and safeguards
Anthropic describes Fable 5 as a Mythos-class model made safe for general use, with capabilities that exceed any model it has ever made generally available. It reports state-of-the-art results on nearly all tested benchmarks — software engineering, knowledge work, vision, and scientific research among them.
The safety design is unusual: rather than refusing outright, queries on some sensitive topics receive a response from Claude Opus 4.8 instead. For now that fallback applies to most requests related to biology and chemistry, and Anthropic says the safeguards trigger in less than 5% of sessions on average.
Pricing and API caveats
Fable 5 costs $10 per 1M input tokens and $50 per 1M output tokens — exactly double Opus 4.8's rate card. Developer-relevant caveats worth knowing before you migrate:
- Thinking is always on — the model reasons on every request; you cannot disable it.
- Refusals surface as a stop reason, not an HTTP error — code that reads content unconditionally will break.
- 30-day data retention is required — organizations on zero-data-retention cannot use it.
Who should use Claude Fable 5
- The hardest reasoning and long-horizon agentic work — the tasks Opus-tier models cannot finish.
- Zero-shot complex builds — early users report strong one-shot results on well-specified systems.
- Deep research and scientific workloads — with the caveat that biology/chemistry requests fall back to Opus 4.8.
At 2x Opus pricing, it is not the default choice. If your workload runs fine on Opus 4.8 or Sonnet 5, the premium buys you little.
How Claude Fable 5 compares
- Claude Fable 5 vs Opus 4.8 — is 2x the price worth it? →
- Claude Opus 4.8 overview →
- GPT-5.6 (Sol) — OpenAI's restricted-preview rival →
Frequently asked questions
When was Claude Fable 5 released?
First on 9 June 2026. Three days later the US Department of Commerce ordered it pulled globally over export-control concerns. Controls were lifted 30 June, and it returned globally from 1 July 2026.
How much does Claude Fable 5 cost?
$10 per million input tokens and $50 per million output tokens — double the rate of Claude Opus 4.8.
What is its context window?
1 million tokens (default and maximum), with up to 128,000 output tokens.
How capable is it?
Anthropic describes it as a Mythos-class model made safe for general use — state-of-the-art on nearly all tested benchmarks, from software engineering to scientific research.
Why do some requests get answered by Opus 4.8?
Safety safeguards route queries on some topics — currently most biology and chemistry requests — to Claude Opus 4.8 instead. This triggers in less than 5% of sessions on average.
Why was it pulled from the market?
On 12 June 2026 the US Department of Commerce instructed Anthropic to pull the model globally over national-security concerns tied to export controls. The controls were lifted on 30 June 2026.