Anthropic launches Claude Fable 5 — subscription billing changes June 23
TL;DR
Anthropic launched Claude Fable 5 (claude-fable-5) on June 9, 2026 — a new Mythos-class tier above Opus 4.8, priced at $10/$50 per million input/output tokens. Pro, Max, Team, and Enterprise subscribers get Fable 5 included through June 22, 2026; from June 23 onward, usage credits are required. A companion invitation-only model, Claude Mythos 5, released simultaneously under the Project Glasswing program.
$10 / $50
Per million input/output tokens — double Claude Opus 4.8's standard pricing of $5/$25
1M tokens
Context window — 128k max output per response, matching extended-context Opus 4.8 configurations
June 22, 2026
Last day Fable 5 is included in Pro/Max/Team/Enterprise subscriptions — usage credits required from June 23 onward
1,000+ hours
Red-teaming prior to launch — Anthropic reports no universal jailbreaks found; 95%+ of sessions run natively on Fable 5
Anthropic launched Claude Fable 5 (model ID claude-fable-5) on June 9, 2026 — its first publicly available model in the new Mythos-class tier, which sits above Claude Opus 4.8 in the lineup. Fable 5 does not replace any existing model; it is an additive release that extends the frontier. A companion model, Claude Mythos 5 (claude-mythos-5), launched simultaneously but is restricted to approved enterprise and security customers under the invitation-only Project Glasswing program.
Pricing doubles Opus 4.8's rate. Fable 5 is priced at $10 per million input tokens and $50 per million output tokens — exactly double the standard Opus 4.8 rate of $5/$25. For a live cross-model price comparison, see the AI Tool Pricing Database. For teams running high-volume agentic pipelines on Opus 4.8 today, this is the most immediate budget impact to model. Cost projections should be re-run before migrating workloads.
The June 22 billing cutoff is an operator action item. All Pro, Max, Team, and Enterprise subscription plans include Fable 5 access through June 22, 2026 at no additional charge. Starting June 23, Fable 5 requires usage credits — it will no longer be covered by the flat subscription. Teams that do not provision credits before that date will either fall back to Opus 4.8 or hit access errors depending on how their integration handles model-unavailable responses. Audit your model routing and credit balance before June 22. TechCrunch's June 9 report noted this billing transition as one of the most operator-relevant changes in the launch.
Hard safety blocks — and what happens when they fire. Fable 5 introduces domain-level hard blocks in cybersecurity, biology, chemistry, and AI-distillation domains. When a request triggers one of these blocks, the model does not error out — it falls back to Claude Opus 4.8 automatically. Anthropic reports that 95%+ of sessions run natively on Fable 5 without triggering a fallback. For the remaining sessions, the fallback is silent from a user-experience standpoint but has billing and performance implications: you are billed at Opus 4.8 rates for those tokens, and the response quality reflects Opus 4.8, not Fable 5. Teams in regulated industries — particularly cybersecurity, pharmaceutical, and chemistry research workflows — should expect some portion of their traffic to hit the fallback and plan accordingly. See our Claude review for how the full model lineup compares on these dimensions.
Safety and red-teaming. Anthropic reports Fable 5 was stress-tested with more than 1,000 hours of red-teaming prior to release, and that no universal jailbreaks were found. The model also carries a mandatory 30-day data retention policy on all Fable 5 traffic — a change from the zero-retention option available on some Opus 4.8 API configurations. Operators in privacy-sensitive verticals should review their data agreements before enabling Fable 5 in production.
Context window and output limits. Fable 5 supports a 1-million-token context window — on par with the extended-context configurations of Opus 4.8 — with a 128k-token maximum output per response. For long-document workflows or large-codebase agents, this matches the upper bound of what Opus 4.8 offered in extended-context mode.
Platform availability. Fable 5 is available on the Claude API, Amazon Bedrock, Vertex AI, Microsoft Foundry, and the Claude Platform on AWS. It surfaces in Claude Code starting with v2.1.170, released the same day. Builders using Claude Code should update their CLI before testing Fable 5 in agentic workflows.
What it means for builders. Fable 5 raises the capability ceiling for teams that can absorb the 2x price increase. For teams weighing Claude against OpenAI alternatives at this new price point, see our Claude vs ChatGPT comparison. The June 23 billing change is the most urgent near-term action item for any team running Fable 5 under a subscription plan. The hard safety blocks require workflow-level testing — particularly for any pipeline touching the four flagged domains — to understand actual fallback rates and how fallback responses affect downstream logic. The Mythos 5 companion model under Project Glasswing remains invitation-only; teams interested in that tier should apply directly to Anthropic.
Why It Matters
The June 23 billing change is the highest-priority action item for any team running Claude on a subscription plan — Fable 5 access drops off subscription coverage without warning if credits are not provisioned. The 2x price increase over Opus 4.8 changes the cost structure for every high-volume pipeline that migrates to the new model. Hard safety blocks in cybersecurity, biology, chemistry, and AI-distillation are a breaking change for workflows in those domains — the silent Opus 4.8 fallback means some requests will receive lower-capability responses without an explicit error, which can be difficult to detect in automated pipelines. The mandatory 30-day data retention on all Fable 5 traffic is a new constraint that may conflict with existing data agreements in privacy-sensitive verticals.
Who's Affected
- — Teams running agentic pipelines on Claude API. The new Mythos-class tier and 2x pricing require re-evaluation of model routing. If you are automating large context window tasks or long-running agents, re-run cost projections against the $10/$50 rate before migrating from Opus 4.8.
- — Subscription users on Pro, Max, Team, or Enterprise plans. You have free access to Fable 5 only through June 22, 2026. Provision usage credits before that date if you intend to continue using Fable 5 — otherwise your integration will fall back to whichever model your code specifies as a default.
- — Cybersecurity, biology, chemistry, and AI-distillation workflows. Hard safety blocks in these domains trigger a silent fallback to Opus 4.8. Test your actual fallback rate before enabling Fable 5 in production. If your pipeline depends on Fable 5-level capability for every response, a 5% fallback rate may be material.
- — Builders evaluating frontier models for regulated or high-stakes applications. The 30-day mandatory data retention policy is new and applies to all Fable 5 traffic. Review your data processing agreements and customer contracts before enabling Fable 5 if you previously relied on zero-retention API configurations.
What To Do Now
- 1. Provision usage credits before June 22. Log into your Anthropic console or AWS/Bedrock/Vertex account and confirm your credit balance covers expected Fable 5 usage beyond that date. Do not wait — the cutoff applies to all subscription tiers simultaneously.
- 2. Audit your model routing for the four hard-blocked domains. Run representative prompts from your cybersecurity, biology, chemistry, or AI-distillation workflows against Fable 5 in a staging environment. Log which requests trigger the Opus 4.8 fallback and decide whether the fallback response quality is acceptable or whether you need to route those requests explicitly.
- 3. Re-run cost projections at the new $10/$50 rate. If you are currently on Opus 4.8 at $5/$25, doubling pricing is material for any high-volume workload. Model the breakeven — do the capability gains justify the price for your specific use case — before committing to a full migration.
- 4. Update Claude Code to v2.1.170 before testing Fable 5 in agentic workflows. Earlier versions do not surface Fable 5. Run `claude --version` to check, then update via your package manager. Test in a non-production environment first to confirm fallback behavior does not affect your agentic pipeline logic.
More on this topic — Best AI Chatbots
Independent Review
Claude
Pricing, pros and cons, real-world verdict — no affiliate spin.
Read the Claude reviewMore from ToolNav News