Fable 5 Leaves Your Plan Tomorrow: Use It at Extra Effort
Last included day is today, credits start tomorrow
Today, July 7, 2026, is the last day Claude Fable 5 is included in your subscription. On Pro, Max, Team, and select Enterprise plans it draws on up to 50 percent of your weekly usage limits at no extra charge. Starting tomorrow, July 8, it moves to usage credits only, billed at $10 per million input tokens and $50 per million output tokens, per Anthropic's own announcement. Anthropic says this is temporary and it aims to bring Fable 5 back to plans when capacity allows.
So this is a timing post, not a pricing complaint. If you are going to spend a chunk of your included allowance today before that window closes, I want to make the case for exactly where to spend it: on Fable 5 at its highest effort setting. Not because higher effort is always worth it, but because Fable 5 is the smarter model and its lead over Opus 4.8 is largest precisely at the top of the effort dial.
Why spend your last included day at Extra effort
Anthropic publishes a chart for its FrontierCode Diamond benchmark, the hardest slice of a long-horizon coding evaluation, that plots each model's score as you turn the effort dial up. I recreated it below from their published figure so you can see the shape of it.
The shape is the whole argument. Opus 4.8 and GPT-5.5 barely move as you spend more effort on them. Fable 5 climbs steadily, from 11.5 percent at low effort to 30.9 percent at its maximum. Fable 5 at merely medium effort, 17.8 percent, already beats Opus 4.8 running flat out at its highest setting, 13.4 percent. In other words, the harder the task and the higher you push the dial, the bigger Fable 5's advantage gets, and today is the last day you can push that dial without paying per token.
This is not only a FrontierCode story. On SWE-bench Pro, Anthropic's headline agentic-coding benchmark, Fable 5 scores 80.3 percent to Opus 4.8's 69.2 percent. Anthropic's own effort documentation puts it plainly: lower effort settings on Fable 5 often exceed the highest effort of prior models. If you have a genuinely hard problem sitting in your backlog, today is the day to throw the smartest available model at it, at full effort, while it is still included.
Extra effort and ultracode are not the same lever
Here is a distinction that took me too long to get straight, because both settings live in the same effort menu and both empty your usage fast.
In claude.ai, the top of Fable 5's effort dial is the setting I have been calling Extra, which is what Anthropic labels xhigh. There is one notch above it, max, which spends noticeably more for a smaller gain. In Claude Code, the comparable top gear for Opus 4.8 is ultracode, and it does something structurally different. Ultracode pairs xhigh effort with standing permission for Claude Code to launch a swarm of subagents on its own, verifying and cross-checking each other without you asking. One is a single model thinking as hard as it can. The other is many agents working the same problem in parallel.
They both burn through a weekly window quickly, and that is where I have to be honest with you. Fable 5 is officially rated to use your limits about twice as fast as Opus per token, and Anthropic warns that ultracode's multi-agent runs consume substantially more tokens than a normal session. In my own use, both empty an allowance in a hurry. But I have not measured them against each other, and I have not found anyone who has published a like-for-like comparison, so I am not going to claim they burn at an identical rate. They do not even burn for the same reason. What I will say is that if you are going to spend a big slice of your allowance on a hard task today regardless, the smarter model is Fable 5, and you get more capability for the tokens you have left by spending them there before tomorrow.
Fable 5 vs Opus 4.8 vs Sonnet 5 after today
Once the included window closes tomorrow, this is the decision you are actually making:
| Model | Included in your plan after today | Cost if metered | Best for right now |
|---|---|---|---|
| Fable 5 | No, usage credits only | $10/M in, $50/M out | Your hardest single tasks, today at full effort |
| Opus 4.8 | Yes, included | $5/M in, $25/M out | Daily driving, most of the work |
| Sonnet 5 | Yes, included | Cheapest of the three | High volume, lower stakes tasks |
The metered figures are Anthropic's own published rates, checked against their pricing page today. Opus 4.8 stays my daily driver from tomorrow. It is included, it is half of Fable 5's per-token cost, and for most of a normal day it is plenty. Fable 5 becomes the model I reach for a couple of times a week on the problems that genuinely earn the credits.
Is Fable 5 gone for good?
No. Anthropic has stated this is a temporary capacity measure, not a removal, and that it plans to restore Fable 5 as a standard part of subscription plans as soon as it can. What is real, starting tomorrow, is that using it beyond your included allowance costs per-token money the same way the API always has.
How to spend your last included Fable 5 day well
If you have not looked at this yet, it takes a few minutes.
- Open claude.ai, go to Settings, then Usage, and check how much of your weekly limit is still available. Fable 5 can draw on up to half of it today.
- Pick the one or two hardest problems you have been putting off, the long, ambiguous, multi-step ones where a smarter model actually changes the outcome.
- Set the effort dial to Extra for those tasks, since that is exactly where Fable 5's lead over Opus 4.8 is largest.
- Switch back to Opus 4.8 or Sonnet 5 for everything routine, so you do not burn through your 50 percent Fable allowance on work that did not need it.
When you hit that 50 percent cap today, Fable 5 pauses until your weekly limit resets, or you switch to another model for the rest of the week. It is not an unlimited last hurrah, so aim it at the work that matters.
What you keep after today
The good news is that the best thing Fable 5 gave me does not depend on Fable 5 access at all. During its free window I had it write six Claude Code skills, the reusable discipline files that make any model plan, verify, and edit more carefully. Those are plain files. They were blind-graded on Opus 4.8 at 12 wins, 0 losses, and 2 ties before any of this, and they run in Codex too. Whatever Fable 5's pricing does next, that discipline is already banked.
You can take the six free skills from the Rigor Pack with no signup and use them on Opus 4.8 from tomorrow. If you want the method behind how they were written, not just the finished files, the Skill Distillation Kit packages the prompts, the blind-test template, and the grading rubric, free and delivered by email.
And if today's reshuffle has you thinking about how much of your setup should not have to be re-decided every time a model changes price or tier, that is the exact problem Iwo's Second Brain is built to solve: a memory layer that stays put while the models underneath it come and go.
FAQ
Did Fable 5's pricing change today? No. Today, July 7, is the last day Fable 5 is included in Pro, Max, Team, and select Enterprise plans, up to 50 percent of your weekly limits. The move to usage credits at $10 per million input and $50 per million output tokens starts tomorrow, July 8.
What is Extra effort on Fable 5? It is the highest standard effort setting, which Anthropic labels xhigh. There is one level above it, max, which costs more for a smaller quality gain. Higher effort means the model thinks longer and uses more tokens.
Is Fable 5 actually smarter than Opus 4.8? On Anthropic's published benchmarks, yes, especially on hard coding and long-horizon tasks. It scores 80.3 percent to Opus 4.8's 69.2 percent on SWE-bench Pro, and on the FrontierCode Diamond effort chart its lead widens the higher you turn up the effort.
Is Extra effort the same as Claude Code's ultracode? No. Extra is a single model, Fable 5, thinking harder on one response. Ultracode pairs Opus 4.8's highest effort with standing permission for Claude Code to launch multiple subagents that verify each other. Same menu, different mechanism, and both use a lot of tokens.
Does the free skills pack stop working once Fable 5 is metered? No. The six skills in the Rigor Pack are static files. They were written by Fable 5 once and run on Opus 4.8, Sonnet 5, or Codex without needing Fable 5 access again.
What happens when I hit my 50 percent Fable 5 limit today? Fable 5 pauses until your weekly usage limit resets. There is no automatic paid overflow while it is still subscription-included, so plan to spend that allowance on your highest-value work first.