Anthropic has decided. Beginning July 20, Claude Fable 5 becomes a permanent part of Max and Team Premium plans at 50% of their limits. Pro and Team Standard users retain access through usage credits and receive a one-time $100 credit. Anthropic officially attributes the staged rollout and repeated extensions to hard-to-predict demand and newly secured capacity. The timing of GPT-5.6 Sol and Kimi K3 makes competitive pressure plausible, but does not prove it caused the decision.
Key takeaways
- ▸From July 20, Fable 5 is permanently included in Max and Team Premium at 50% of plan limits.
- ▸Pro and Team Standard move to usage credits and receive a one-time $100 credit.
- ▸Anthropic says demand was difficult to predict and access expanded as it secured more inference capacity.
- ▸Kimi K3 launched four days before the announcement with much lower API prices, but Moonshot itself says K3 still trails Fable 5 and Sol overall.
- ▸Our inference: capacity explains the rollout; competition likely raised the commercial cost of removing Fable 5. Anthropic has not confirmed that link.
In this article
1. What changes on July 20
The full timeline, because at this pace it's easy to lose track:
| Date | Event |
|---|---|
| June 12, 2026 | US export controls: Anthropic cuts off worldwide access to Fable 5 and Mythos 5. |
| June 29-30, 2026 | The Commerce Department lifts the controls; Anthropic announces the redeployment. |
| July 1, 2026 | Fable 5 becomes available worldwide again and is included in paid plans (~50% of weekly limits), with an announced deadline of July 7. |
| July 7, 2026 | First extension, announced hours before the cutoff after a wave of user protests: the window now runs through July 12. |
| July 9, 2026 | An outside event, but a decisive one: OpenAI ships GPT-5.6 Sol, Terra and Luna. Sol positions itself head-on against Fable 5, at half the price per million tokens. |
| July 12, 2026 | Second extension, announced on the very day of the deadline: the window now runs through July 19 (11:59 PM Pacific Time), and Claude Code keeps its weekly limits raised by 50%. |
| From July 20 | Permanent inclusion at 50% of limits for Max and Team Premium. Pro and Team Standard use credits and receive a one-time $100 credit. |
Six different access regimes for the same model in thirty days. Anthropic maintains its careful wording: the move to usage credits is not presented as a permanent removal from subscriptions.
2. Why Anthropic changed the model
Anthropic's official explanation is operational: demand for Fable 5 was difficult to predict, so access was extended in stages as the company secured additional capacity. The successive deadlines were therefore capacity checkpoints, not a settled long-term product policy.
The permanent arrangement keeps that constraint visible. Max and Team Premium receive Fable 5 at 50% of their limits, while Pro and Team Standard stay on usage credits. The one-time $100 credit cushions the transition without promising unlimited inference inside lower-priced subscriptions.
Our July 12 prediction was a permanent inclusion in higher plans with a reduced cap, combined with usage credits elsewhere. That is essentially what Anthropic chose, with Team Premium included alongside Max. It validates the economic diagnosis; it does not prove every hypothesis about the forces that influenced the decision.
3. Competitive pressure from Sol and Kimi K3
Capacity is the only cause Anthropic states. The market context nevertheless matters: GPT-5.6 Sol and Kimi K3 give buyers credible alternatives, while Anthropic must protect both retention and inference margins. The following is therefore a competitive analysis, not a claim about Anthropic's internal decision-making.
First jaw: GPT-5.6 Sol, nearly the same measured intelligence at half the price
Since July 9, OpenAI's commercial argument fits on one line: on the Artificial Analysis Intelligence Index, Sol measures 59 where Fable 5 measures 60 — and Sol costs $5 / $30 per million tokens where Fable 5, on usage credits, will cost $10 / $50. Measured per task, the gap reaches roughly a factor of three, and on the coding agent index Sol even pulls ahead at about 40% less per task. The picture is not one-sided — our analysis of the release details the benchmarks where Fable 5 keeps a clear lead (over 80% on SWE-bench Pro versus 64.6% for Sol) and what Sol's system card documents about overstepping instructions. But for a buyer comparing two price lines, the pressure is real: every week Fable 5 stays included in subscriptions is a week that comparison doesn't get made.
Second jaw: free Chinese models at frontier level
The other front sells nothing at all. In late June, Zhipu released GLM-5.2 as open source with no usage restrictions, and MiniMax opened the weights of its flagship M3 model, reported at GPT-5.5 level in coding and agentic reasoning (CNBC, June 2026). On OpenRouter, the largest neutral model router, about 61% of all tokens consumed were already flowing through Chinese open-weight models in early June. Against a free alternative improving at this pace, every dollar of premium pricing has to be bought with a measurable capability gap — and that gap narrows with every release. It is the thesis we developed in our analysis of the two competing visions for AI (in French) — and every episode of the Fable 5 saga confirms it a little more.
Third jaw — the tightest: Anthropic's own IPO
Anthropic filed for its initial public offering on June 1, targeting a Nasdaq listing in October at a valuation of about $965 billion — potentially the first IPO in history to cross the trillion mark. Three months from a listing, the double bind becomes arithmetic: you present investors neither a user base migrating to a competitor at half the price, nor a margin eroded by giving away inference on the most expensive model on the market. The weekly extension, renewed at the last minute, is the exact trace of that unresolved trade-off. After October, it will no longer be settled internally: it will be settled in front of a stock price, quarter after quarter.
Why this matters to a European SME
The price and access conditions of the best model on the market are decided in a Nasdaq–Hong Kong–Washington triangle, light-years from your annual budget — and they have changed six times in thirty days. It is the thesis our founder defended in his op-ed for Journal du Net (in French): the model is a commodity whose terms change without you. Your durable advantage lives in your data, your processes and the guardrails you impose — not in betting on a logo.
4. What users should do now
The deadline is gone, but the economic choice is clearer. Three actions, in order:
1. Calculate before upgrading from Pro to Max
Use the $100 credit on your hardest real tasks. Measure cost per completed case and compare it with the annual premium for Max or Team Premium.
2. Add Sol and Kimi K3 to the test bench
Run the same cases with the same tools and criteria. Fable 5 may justify its premium where it reduces human rework; Sol or Kimi K3 may transform the economics of volume tasks.
3. Make your processes indifferent to the next announcement
Six access regimes in just over a month show that no critical process should depend on one model. Task routing, regression tests and automatic fallback remain useful whatever comes next. They are central to our approach to enterprise AI automation.
5. FAQ
Is Claude Fable 5 now permanently included?
Yes, in Max and Team Premium from July 20, 2026, at 50% of plan limits. Anthropic announced no new end date. Pro and Team Standard retain access through usage credits.
What do Pro and Team Standard users receive?
They receive a one-time $100 credit and continue to access Fable 5 through usage credits. Anthropic has not announced recurring included Fable usage for those plans.
Why did Anthropic extend access several times?
Anthropic says demand was difficult to predict and access expanded in stages as additional capacity became available. It does not name GPT-5.6 Sol or Kimi K3 as causes.
Is GPT-5.6 Sol better than Claude Fable 5?
It depends on the task — single leaderboards no longer exist. On the Artificial Analysis Intelligence Index, Sol measures 59 against 60 for Fable 5, and it leads the coding agent index at about 40% less per task. But Fable 5 exceeds 80% on SWE-bench Pro against 64.6% for Sol, and GPT-5.6's system card documents a stronger tendency to overstep instructions. Test on your real cases before concluding.
Did Kimi K3 force Anthropic's decision?
No public evidence proves it. Kimi K3 arrived four days before the announcement with much lower prices and competitive results on selected tasks. It is plausible competitive pressure, not an established cause.
Closing thoughts
Anthropic has replaced repeated extensions with a clear rule: Fable 5 is a permanent benefit of Max and Team Premium, under a reduced quota, while Pro and Team Standard use credits. Capacity is the official reason. Sol and Kimi K3 make competitive pressure plausible, not proven. For businesses, the durable response is to measure cost per result, keep several models tested and design processes that survive the next pricing or quota change.
Would your AI budget survive six price changes in a month?
30 minutes to map your model dependencies, quantify where the premium tier pays off, and set up routing that absorbs the announcements for you.
Sources
- Official permanent Fable 5 plan announcement — @claudeai on X (07/18/2026)
- Kimi K3: Open Frontier Intelligence — Moonshot AI (07/14/2026)
- Official announcement of the July 19 extension — @claudeai on X (07/12/2026)
- Anthropic extends Claude Fable 5's access on paid plans until 19th July — OfficeChai (07/12/2026), including the Artificial Analysis comparisons
- Claude Fable 5 promotion extended after backlash over early cutoff — Android Authority (07/07/2026)
- China's Zhipu is closing in on top U.S. AI models — CNBC (06/26/2026)
- Zhipu surges 33% as Wall Street raises bets on China AI — CNBC (06/15/2026)
- Redeploying Claude Fable 5 — Anthropic
The IPO and valuation figures (S-1 filed June 1, ~$965bn, Nasdaq listing targeted for October) are detailed and sourced in our analysis of the China / US AI war (in French).
Victor Gless-Krumhorn
Founder & AI Consultant — JAIKIN
AI implementation and automation expert for SMBs and mid-market companies. Works with businesses across France, Germany and Switzerland, from process mapping to production rollout.