Claude Fable5 Free Trial & Free Access (Updated July 2026)
Every legit route to free or cheap Claude Fable5 access after the July 1 restoration: the included window through July 7, the API starter credit, and caching tricks that cut costs by up to 95%.
Fable5 Editorial
Update — July 1, 2026: Fable 5 is back worldwide — restored on Claude.ai, Claude Code, and the API (
claude-fable-5) after the US export-control directive was withdrawn on June 30. This guide applies directly again. Note: the original June 22 free window has passed; the current rule is that Pro/Max/Team plans include Fable 5 in up to 50% of the weekly usage cap through July 7, 2026, then it moves to usage credits ($10/$50 per million tokens). Live status: /is-fable-5-back · What changed: Claude Fable 5 is back.
The honest answer first: there is one genuinely free route to Claude Fable5, and it now expires on July 7, 2026 — the post-restoration included window that replaced the original June 22 one. Everything else is about making paid access dramatically cheaper. This guide covers every legitimate option — what's free now, what changes on July 8, and how to stretch every dollar after that. No promo codes exist; anyone selling "Fable5 keys" outside official channels is scamming you.
Want the live answer? Our status page tracks the included window in real time and mirrors the official service status.
Route 1: The included subscription window (actually free)
When Anthropic launched Fable5 on June 9, it included the model at no extra cost on Pro, Max, Team, and seat-based Enterprise plans through June 22. The June 12 suspension froze that plan; after the July 1 restoration, Anthropic reset it: through July 7, 2026, Fable5 is included in up to 50% of your weekly usage cap. If you already pay for any of those plans, you have Fable5 right now:
- Sign in to the Claude app — web, desktop, or mobile — with the account on your paid plan.
- Start a new conversation.
- Open the model picker (top of the chat, near the message box).
- Select Claude Fable5.
- Send your prompt. You're running the Mythos-class model without per-token charges.
Normal plan usage limits still apply — Fable5 counts against at most 50% of your weekly cap, and it is token-hungry, so a Pro plan's allowance goes fast. Field reports describe burning a day's Max allowance in a single long agentic run. Use the window for the tasks that deserve it: hand Fable5 a whole job (a feature build, a deep code review, a research synthesis), not small interactive asks. Our prompt guide covers how to frame those one-shot jobs.
In Claude Code, switch with /model claude-fable-5 — same window, same plan limits.
What changes on July 8
Anthropic is explicit about the cliff:
- July 8: the included window ends. Using Fable5 on a subscription requires usage credits, metered at API rates ($10 in / $50 out per million tokens).
- Standard Enterprise seats have no included allowance at all — credits only, from day one.
- One more post-restoration note: the new safety classifiers occasionally reroute routine coding/debugging requests to Opus 4.8 while Anthropic tunes false positives — fallback responses bill at Opus rates.
A related deadline already passed on June 15: Agent SDK usage (claude -p, Claude Code GitHub Actions, third-party SDK apps) moved to a separate monthly credit pool — $20 on Pro, $100 on Max 5x, $200 on Max 20x — billed at API rates. Interactive chat and terminal Claude Code are unaffected, but if your automation runs headless, it's already metered.
Route 2: The API starter credit (~$5, one-time)
The Claude API has no ongoing free tier, but new accounts at the Claude Console receive a one-time starter credit of roughly $5, no credit card required. That works with any model, including claude-fable-5.
Five dollars sounds like nothing against $10/$50 pricing — but it's enough for a real evaluation if you spend it carefully:
- A focused test with a 5K-token prompt and 2K-token response costs about $0.15. That's ~30 meaningful test calls.
- Watch the
usageobject on your first response so you know your real per-call cost before committing. - Don't burn the credit on long agentic runs — those are exactly what eats tokens fastest.
Route 3: Channels you may already pay for
Not free, but possibly already covered by something you subscribe to:
- GitHub Copilot — Fable5 is generally available for Copilot Pro+, Max, Business, and Enterprise (not Free or base Pro). It bills at provider list pricing under usage-based billing, and org admins must enable the Fable5 policy first (off by default, because the model requires 30-day data retention).
- OpenRouter — listed as
anthropic/claude-fable-5. Same list price, but provider-level caching means the effective input price has been averaging around $6.40 per million. Useful if you already hold OpenRouter credits. - AWS Bedrock / Google Cloud / Microsoft Foundry — enterprise channels; if your company has committed cloud spend, Fable5 may be reachable through an existing contract rather than a new bill. (Post-restoration these are being re-enabled "as fast as possible" — check your console.)
Stretching paid usage: the 95% stack
Once you're paying, three mechanisms compound:
| Lever | Discount | Fable5 price after |
|---|---|---|
| Base input | — | $10 / MTok |
| Prompt caching (hits) | −90% | $1 / MTok |
| Batch API | −50% on everything | $5 in / $25 out |
| Caching + Batch stacked | ≈ −95% on cached input | ≈ $0.50 / MTok |
Practical translation:
- Cache your system prompt and tool definitions. First write costs 1.25x ($12.50/MTok for 5-minute TTL, $20 for 1-hour), every subsequent hit costs $1/MTok. Long-running agents amortize this almost immediately.
- Batch anything that can wait. The Batch API takes up to 24 hours (most batches finish within one) and halves both input and output. Evaluations, overnight processing, bulk generation — all batch material.
- Stack them. Cached batch input lands around 5% of list price. Anthropic confirms the discounts combine.
One more quirk in your favor: when Fable5's safeguards reroute a request to Opus 4.8, the response bills at Opus rates ($5/$25) — you're never charged Fable prices for fallback answers.
Run your own numbers in our cost calculator — it compares Fable5 against Opus 4.8, Sonnet 4.6, and Haiku 4.5 for your actual monthly volume.
The decision tree
- Have a paid Claude plan? Use Fable5 in the apps through July 7 — included in up to 50% of your weekly cap. Today.
- Developer without a plan? Create a Console account, claim the ~$5 starter credit, run a controlled evaluation.
- Already on Copilot Pro+/Business or holding OpenRouter credits? Check those pickers before paying Anthropic directly.
- Going to production? Cache + batch + route: Fable5 for the frontier-hard 10–20%, Opus 4.8 for the rest.
Dated June 10, 2026; updated July 2, 2026 after the restoration. The July 7 deadline and credit amounts are Anthropic's published figures. Planning ahead? Here's how to keep using Fable5 after the included window.