July 7, 2026. If your team leans on Claude Fable 5, today is the day that matters. Today is the last day Fable 5 is included in Pro, Max, Team, and premium Enterprise plans at up to 50% of weekly usage limits. Starting tomorrow, July 8, keeping access means enabling usage credits, a metered billing layer on top of your plan, at Anthropic's highest published rate for a generally available model. This is the second time Fable 5 has moved from bundled to metered since its June 9 launch (the export-control shutdown reset the first window), and this time the switch sticks until Anthropic can build enough compute to serve it. Nothing else in your plan changes: Opus 4.8, Sonnet 5, and Haiku stay included. But Fable 5 becomes a line item you have to turn on, or it stops.
The key developments
- The deadline is today. Per Anthropic's redeployment announcement, the included Fable 5 allowance (up to 50% of weekly limits since the July 1 return) ends July 7. From July 8, all Fable 5 sessions draw from a separate usage-credit balance.
- The rate is steep. The Claude pricing page lists Fable 5 at $10 per million input tokens and $50 per million output tokens, exactly double Claude Opus 4.8 at $5/$25 and the most expensive rate Anthropic has set for a widely available model. On real workloads the output rate dominates: a pass that reads 200,000 tokens and writes 40,000 costs about $4.00 on Fable 5 versus roughly $0.80 on Sonnet 5.
- No credits enabled means no access. You must turn on usage credits and set a spending cap in Settings then Usage on claude.ai (or the Console) before your allowance runs out. There is no automatic fallback: if the balance empties, Fable 5 simply stops, mid-session, with no grace period.
- Who gets hit. Standard Enterprise seats never had Fable 5 included, so nothing changes there. The change lands on Pro, Max, Team, and premium Enterprise users, and hardest on developers running Fable 5 inside agentic loops, where the full context is re-billed as input on every turn.
- Anthropic calls it temporary. The company frames usage credits as a way to manage demand it expects to be very high and hard to predict, not a permanent price tier, and says it will return Fable 5 to standard subscriptions when capacity allows. No date has been given.
What it means for operators
Treat today as a cost-control checkpoint, not a scramble to keep Fable 5 on by default. First, decide what genuinely needs it: its 1-million-token context and long-horizon coherence earn the premium on large code migrations and multi-day autonomous work, and little else. Everything else should route to Claude Sonnet 5 at its $2/$10 introductory rate or to Opus 4.8. Second, if you do keep it, pull the cost levers: prompt caching cuts repeated input by 90% (from $10 to $1 per million), and the Batch API halves both rates for anything that is not real time. Third, log which model actually answered each request, because the post-relaunch safety classifier reroutes some coding tasks to Opus 4.8, which changes both cost and the model you thought you were using. The broader signal is the one we flagged when ChatGPT agents went metered and when the whole business AI stack turned its meters on: frontier access is now billed by the token, so the winning posture is a model-agnostic automation setup with caps and routing from day one, not a single model wired in as if its price will never move. If you would rather that discipline be built and monitored for you, that is exactly what an AI engineer who tracks cost per outcome delivers.
Frequently Asked Questions
July 7 is the last day Fable 5 is included in Pro, Max, Team, and premium Enterprise plans (up to 50% of weekly usage limits). From July 8, Fable 5 no longer draws from your subscription and instead bills through usage credits at API rates. The rest of your plan, including Opus 4.8, Sonnet 5, and Haiku, is unaffected.
Fable 5 bills at $10 per million input tokens and $50 per million output tokens, double Opus 4.8's $5/$25 and Anthropic's highest rate for a generally available model. Prompt caching cuts repeated input by 90% (to $1 per million), and the Batch API halves both rates for work that does not need a real-time response.
Access stops. There is no automatic fallback to another model, so a running session ends when the included allowance or credit balance empties. Anthropic directs users to enable credits and set a monthly spending cap in Settings then Usage on claude.ai (or the Console) before the cutoff.
Route by task rather than defaulting everything to it. Reserve Fable 5 for work that genuinely needs its 1-million-token context and long-horizon coherence, such as large code migrations or multi-day autonomous agents, and send everything else to Sonnet 5 (introductory $2/$10 through August 31) or Opus 4.8. Log which model answered each request, since the safety classifier reroutes some coding tasks to Opus 4.8.