Claude Fable 5 API Access: Pricing, US Restrictions & International Alternatives (2026)

The access reality as of June 2026

Fable 5 launched June 9, 2026 as Anthropic's first public Mythos-class model — the most capable generally available AI model released to date, scoring 80.3% on SWE-Bench Pro and 59% on Humanity's Last Exam. Three days later, on June 12, a jailbreak technique bypassed Fable 5's safety constraints and exposed the underlying Mythos 5 model's autonomous cybersecurity capabilities. The US Department of Commerce issued an emergency export control directive the same day. Anthropic shut both models globally within 90 minutes. International users lost access not because of a payment tier but because the export control applies by nationality — and Anthropic had no technical means to filter users that quickly at model invocation.

The practical result: Fable 5 is US-only, and the access check is server-side. Routing through a US-based proxy or VPN does not help — the check runs at model invocation, not at login or billing. Accounts with verified US-person status can call `claude-fable-5` normally; all others receive an access-denied response regardless of plan or payment method.

Fable 5 pricing for US accounts

Anthropic included Fable 5 at no extra cost in all paid plans — Pro, Max, Team, and Enterprise — through June 22, 2026. From June 23 onward, usage draws on credits at $10 per million input tokens and $50 per million output tokens. Context caching and batch API discounts apply at the same structure as Claude Opus 4.8 and other models. For automation workloads, the metered API at this rate is the correct billing surface; for all-day interactive coding or analysis, a Claude Max plan caps spend in a way the per-token rate never will. The subscription-vs-API breakeven applies directly: if a day's Fable 5 API tokens would cost more than the plan's daily amortized cost, the subscription wins.

Is there a free Fable 5 tier?

No. The claude.ai chat app runs a capable model on the free tier, but Fable 5 is not available there — not even for US users on the free tier. The chat app's free tier runs older models at daily caps; Fable 5 requires a paid plan (and, for now, verified US-person status). There is no free API tier for Fable 5, no trial credits specifically for it, and no third-party aggregator that can front it legally for international users under the current export controls.

Best alternatives for international teams

The substitution decision turns on how close to Fable 5's capability ceiling a given workload actually needs to run. For the majority of enterprise workloads that were running on Opus 4.8 before the Fable 5 launch, the rollback is clean — Opus 4.8 is internationally available, has a 200K context window, and scores 69.2% on SWE-Bench Pro agentic coding. The 11-point gap to Fable 5 (80.3%) is real and large, but most real-world tasks do not live at that ceiling.

For teams that had moved workflows specifically to leverage Fable 5's agentic coding ceiling or long-context capability, GPT-5.5 from OpenAI is the nearest cross-vendor option — no comparable export restriction was in place as of June 2026. Llama 4 Maverick is the strongest open-weights route: self-hostable, no nationality restriction, and near-frontier on most benchmarks, at the cost of GPU spend and operational overhead.

Access summary by profile

US-verified account, interactive use: Claude Max plan, Fable 5 included. US-verified account, automation: direct API at `claude-fable-5`, $10/$50 per million tokens. International team, Anthropic continuity: Claude Opus 4.8 globally available. International team, closest to Fable 5 capability: GPT-5.5 (OpenAI, no comparable export restriction). International team, sovereignty-safe: Llama 4 Maverick self-hosted. The export controls explainer covers the legal and procurement implications in detail. Fable 5 access status last verified 2026-06-20; this is a fast-moving situation.

Is Claude Fable 5 free?

No. Fable 5 was included at no extra cost in Claude Pro, Max, Team, and Enterprise plans through June 22, 2026. From June 23 onward, it draws on usage credits at $10/M input tokens and $50/M output tokens. There is no permanent free tier for Fable 5, and there is no free API access. For international users, the question is moot: Fable 5 is blocked regardless of payment tier under US export controls in effect since June 12, 2026.

Can international users access Claude Fable 5?

No. On June 12, 2026, the US Department of Commerce issued an export control directive suspending Fable 5 and Mythos 5 access for all foreign nationals, including non-citizen Anthropic staff. Anthropic shut both models globally within 90 minutes — it could not filter users by nationality in real time across its cloud platforms. The restriction is by verified nationality, not by billing address or proxy location: the check runs server-side at model invocation. The timeline for lifting the restriction is unknown; Anthropic and Commerce are in active negotiation, with a multi-month resolution expected.

What does the Claude Fable 5 API cost?

The Fable 5 API (model ID: `claude-fable-5`) is priced at $10 per million input tokens and $50 per million output tokens. Context caching and batch API discounts apply at the same structure as other Claude models. Through June 22, 2026, paid plan subscribers could run Fable 5 against their plan's usage allowance at no extra cost; that transitional period ends June 23. The API is only accessible to accounts with verified US-person status.

What is the best alternative to Fable 5 for international teams?

Within the Anthropic family, Claude Opus 4.8 is the strongest available model for international users: 69.2% on SWE-Bench Pro agentic coding (vs Fable 5's 80.3%), 200K context, available globally. For teams that need a model closer to Fable 5's capability ceiling, GPT-5.5 from OpenAI is the nearest cross-vendor option (no comparable export restriction as of June 2026) and Llama 4 Maverick is the strongest open-weights route (self-hostable, no nationality restriction). The performance gap to Fable 5 is real — Humanity's Last Exam put Fable 5 at 59% vs GPT-5.5's 41.4% — so the substitution depends on how hard the ceiling matters for your workload.

Will Claude Fable 5 access be restored internationally?

Anthropic has committed to building nationality-verification infrastructure that would allow selective re-activation — serving US-eligible users while blocking foreign nationals — but that infrastructure does not yet exist. Commerce and Anthropic are in active negotiation. No public timeline has been given; industry analysts and legal commentators are working from a multi-month estimate. The DoD has already migrated two-thirds of its Fable 5 workflows to alternatives, which reduces the commercial pressure on a fast resolution.

Is there any free route to Fable 5 capability for non-US teams?

No legitimate free route. VPN and proxy workarounds fail at model invocation — the nationality check is server-side. The honest answer is that there is no free, legal path to Fable 5 for international users while the export controls hold. The practical decision is which alternative to route to: Opus 4.8 for Anthropic-API continuity, GPT-5.5 for the best-available cap, or Llama 4 Maverick for a self-hosted, sovereignty-safe option.