Updated 18 July 2026: The June free window is over, but the news is better than expected. Fable 5 is now included in all Claude Max and Team plans, and Pro subscribers get a one-time $100 credit to try it, after which it runs on pay-as-you-go usage. The sections below are updated for this.
TL;DR: Claude Fable 5 is Anthropic’s most powerful publicly available AI model. It is now included in every Claude Max and Team plan at no extra cost. On the Pro plan, Anthropic gives you a one-time $100 credit to try it, and after that credit runs out you pay only for what you use. I have been running it inside Claude Code on my own projects since launch, and it is genuinely the biggest jump I have felt since Opus launched.
I rarely write about an AI model within days of its launch. Most “new model” releases feel like a 5% improvement with 100% more marketing.
This one is different.
Anthropic released Claude Fable 5 on 9 June 2026, and I switched my Claude Code sessions to it the same week. Since then, I have used it to design and build work on several complex projects I own — a Next.js site revamp, a school directory platform I am building, and a couple of my WordPress plugins. It has handled all of them amazingly well.
The best part? If you are on Max or Team, it is simply part of your plan now. If you are on Pro, Anthropic hands you $100 of credit to see whether it is worth it. Here is everything you need to know.
What Is Claude Fable 5?
Short answer: it is the first model of Anthropic’s new Claude 5 generation, and it sits above Claude Opus in capability.
Anthropic calls it a “Mythos-class” model that has been made safe for general use. The unrestricted version, Claude Mythos 5, is limited to vetted research partners. Fable 5 is the version you and I actually get — same intelligence, with extra safety guardrails on sensitive topics like cybersecurity and biology.
The headline specs, straight from Anthropic’s official model page:
- 1 million token context window — it can hold an entire large codebase or several books in one conversation.
- Built for long-running work — Anthropic says it can run agentic tasks for days at a time, planning across stages and checking its own work.
- Thinking is always on — the model reasons before every answer. You cannot switch it off, and honestly, you would not want to.
- API pricing of $10 per million input tokens and $50 per million output tokens — roughly double the Opus tier, which tells you where Anthropic places it. This is also the rate your Pro credit and pay-as-you-go usage are measured against.
How Good Is Fable 5 Compared to Opus 4.8 and GPT-5.5?
On Anthropic’s published benchmarks, Fable 5 leads Claude Opus 4.8, GPT-5.5, and Gemini 3.1 Pro on most rows — agentic coding, knowledge work, multidisciplinary reasoning, and computer use.
Benchmarks are benchmarks. The early customer reports are more interesting. Stripe’s engineering team said Fable 5 did in a day what would have taken more than two months by hand in their 50-million-line codebase. Cursor’s CEO called it the state-of-the-art model on their internal benchmark.
My own experience matches that direction. More on it below.
How Do You Get Claude Fable 5 on Your Plan?
It depends on which Claude plan you are on. There are two paths now:
- Max ($100 or $200/month) and Team plans — included. Fable 5 is part of the plan at no extra cost. Nothing to buy, nothing to activate. Your normal plan usage limits apply, exactly like any other model.
- Pro ($20/month) — $100 one-time credit, then pay-as-you-go. Anthropic gives Pro subscribers a one-time $100 credit to try Fable 5. Once that credit is used up, you keep access on a pay-as-you-go basis — you top up usage credits and pay only for what you actually consume.
Either way there is no separate sign-up, no waitlist, and no coupon code. You simply select the model:
- Open claude.ai (web, desktop, or mobile app), or open Claude Code in your terminal.
- Click the model selector (in Claude Code, type
/model). - Choose Claude Fable 5 from the list.
- Start working. That is it.
This is a much better place than where we were in June. At launch, Fable 5 was free for everyone for two weeks and then pulled out of subscriptions entirely on 23 June because of compute capacity. Anthropic has now brought it back properly — permanently included on Max and Team, and reachable on Pro through the $100 credit and usage-based top-ups.
How Do You Run Fable 5 in Claude Code?
This is where Fable 5 truly shines. The chat app is nice, but Claude Code is where this model’s long-horizon ability actually shows up — it is an agent that works in your terminal, reads your codebase, edits files, runs tests, and keeps going till the task is done.
Setting it up takes under two minutes:
- Install Claude Code. Run
npm install -g @anthropic-ai/claude-code, or copy the one-line install command from the official Claude Code docs. If the terminal feels new to you, my Mac Terminal basics guide covers the first steps. - Sign in with your Claude subscription. Run
claudeinside any project folder and log in with your claude.ai account — not an API key. - Switch the model. Type
/modeland select Fable 5 from the list. - Give it real work. Not “write a function”. Give it a full feature, a migration, or a messy bug. That is where the difference from Opus shows.
On Max and Team, your normal plan usage limits still apply — included does not mean unlimited. On Pro, keep an eye on how fast the $100 credit moves, because Fable 5 thinks before every answer and long agentic sessions consume more than you expect.
My Experience After Real Work
Let me be honest: I expected another incremental upgrade. That is not what this is.
I have thrown my actual work at Fable 5 inside Claude Code — not toy prompts. A Next.js site revamp with custom Markdoc components. A school directory platform I am building with search, reviews, and an admin panel. Plugin work on my WordPress products.
Three things stood out:
- It holds the whole project in its head. On my directory platform, it traced a bug across the API layer, the cache, and the frontend in one go. Older models would fix one layer and break another.
- It designs before it codes. Ask it to build a feature and it plans the structure, edge cases, and tests first. The first version it ships is usually the version that works.
- It checks its own work. It runs the build, reads the errors, and fixes them without me pasting anything back. I review the result instead of supervising every step.
My take: Fable 5 is the first model where I hand over a complex task, walk away to make chai, and come back to work I would actually merge. It works amazingly well on long, messy, multi-file tasks — exactly the work I used to break into small pieces by hand.
It is not magic. It thinks longer before answering, so quick one-line questions feel slower than Opus. For those, I still switch back. But for serious project work, it has been my default since 10 June.
What Is the Catch?
There are three, and you should know them before you depend on it.
- On Pro, the $100 credit is one-time. After it is used up you are on pay-as-you-go, and at $10/$50 per million tokens, heavy daily use adds up quickly. If Fable 5 becomes your main working model, Max works out cheaper than topping up Pro every month.
- Some queries get answered by Opus 4.8 instead. Anthropic added safety classifiers for cybersecurity and biology topics. Flagged queries are quietly routed to Opus 4.8. Anthropic says this triggers in under 5% of sessions, and across my web and coding work I have never hit it.
- 30-day data retention is mandatory. Fable 5 is not available under zero-data-retention setups because Anthropic keeps conversations for 30 days for safety monitoring. For my blogging and development work this changes nothing, but teams with strict data policies should check first.
None of these is a dealbreaker for a solo builder. The one that decides your plan is the first — how much of the model you actually intend to use.
FAQ
Is Claude Fable 5 free?
On Max and Team plans, yes — it is included at no extra cost, within your normal plan usage limits. On Pro, you get a one-time $100 credit to try it, and after that you pay for usage. There is no access on claude.ai’s free (no-subscription) tier.
What is the $100 Pro credit, exactly?
It is a one-time usage credit Anthropic gives Pro subscribers so they can try Fable 5 without committing to a bigger plan. You use it like any usage credit — it draws down as you send tokens through the model. Once it is finished, Fable 5 stays available on pay-as-you-go, so you top up and pay only for what you use.
I thought Fable 5 was removed from subscriptions on 23 June. What changed?
That was the position right after launch, when Anthropic pulled Fable 5 out of subscription plans because of compute capacity. It has since come back — now included in all Max and Team plans, and available on Pro through the $100 credit plus pay-as-you-go usage.
Can I use Fable 5 with the $20 Claude Pro plan?
Yes, through the $100 credit and then pay-as-you-go, in both the Claude apps and Claude Code. Just remember Fable 5’s longer thinking consumes credit faster than Opus does, so a few long agentic sessions will move that balance quicker than you expect.
Is Fable 5 better than Claude Opus 4.8?
For long, complex tasks — yes, clearly. Anthropic’s benchmarks show it ahead on agentic coding, reasoning, and knowledge work, and my own use matches that. For quick, simple questions, Opus 4.8 still feels snappier — and on Pro, it does not spend your credit.
What is the difference between Fable 5 and Mythos 5?
Same underlying model. Fable 5 is the general-release version with safety classifiers on cybersecurity and biology topics. Mythos 5 removes those measures and is available only to approved organisations and vetted researchers.
Does Fable 5 work in Claude Code on Windows?
Yes. Claude Code runs on macOS, Linux, and Windows, and Fable 5 is a selectable model on all of them. You can also use it through the Claude Code desktop app, the VS Code and JetBrains extensions, or the web.
Summing Up!
Claude Fable 5 is the rare model launch that lives up to its own announcement. After building my own projects with it in Claude Code, I can say the jump over Opus is real — especially for long, complex, multi-file work.
My recommendation is simple. If you are on Max or Team, you already have it — open Claude Code today, type /model, pick Fable 5, and give it your hardest pending task. If you are on Pro, spend that $100 credit on one genuinely hard project rather than small prompts, and you will know within a week whether Max is worth it for you.
If you are exploring AI agents more seriously, my write-up on Grip AI, my open-source AI agent platform, pairs well with this — and if you are a student hunting for free AI tools, check whether the Google AI Pro student offer is live for you.
Have you tried Fable 5 yet — and what is the biggest task you have handed to it? Tell me in the comments.