How to Use Claude Fable 5: A Practical Getting Started Guide

Claude Fable 5 came out on June 9, 2026, and it's Anthropic's first publicly available Mythos-class model. If you've heard the name and just want to start using it, this guide keeps things practical. There are really two paths depending on who you are: open one of the Claude apps and pick the model, or call the API with a model id. I'll walk through both, then cover what the model is actually good at so you don't waste tokens on the wrong jobs.

One thing up front. ShadowProof is independent and we're not affiliated with Anthropic. We make a US mobile IP app for creators, and we write about tools our readers ask about. So treat this as a straight how-to, no marketing attached.

Using Claude Fable 5 in the Claude apps

The easiest way in is through the Claude apps and claude.com. If you're on a Pro, Max, Team, or seat based Enterprise plan, you've got access without doing anything special on the billing side.

Pick the model and start a chat

The general flow is the same one you already know. Start a new conversation, then choose Claude Fable 5 from the model picker before you send your message. I'm not going to invent exact button labels or screenshots here, because interfaces change and I'd rather you trust this page. The short version: look for the model selector in your chat window and switch it to Fable 5. From there you just type like normal.

If you don't see it as an option, check that your plan is one of the eligible ones above and that your app is up to date.

The free window, and what happens after

Here's the part worth circling. From June 9 to June 22, 2026, Fable 5 is included in those Pro, Max, Team, and seat based Enterprise plans at no extra cost. You can use it during that stretch without burning anything separate.

After June 22, using Fable 5 on those plans needs usage credits. So if you've been leaning on it heavily, that's the date to keep in mind. One honest caveat: this window is time limited, and Anthropic can change the dates or terms. Don't build a workflow that assumes free access lasts forever. If you want the cost breakdown for after the window closes, our Claude Fable 5 pricing guide walks through the numbers.

Using Claude Fable 5 through the API

If you're building something, the API is where you'll live. The model id is the whole trick.

The model id

Pass 'claude-fable-5' as the model in your API request. That's it. Wherever your code or SDK names the model, swap in claude-fable-5 and your calls route to the new model. If you've already got Claude API code running against another model, this is usually a one line change to point at the new id.

A quick reminder on cost, since the API bills by token and not by subscription: it's $10 per million input tokens and $50 per million output tokens. Those output tokens are the expensive ones, so if you're feeding huge inputs and asking for short answers, the math works in your favor. If you're generating long outputs at volume, watch the meter.

Other ways to reach it

The API isn't the only programmatic option. You can also get Fable 5 through:

• Consumption based Enterprise plans, billed by usage.
• AWS Bedrock, if your stack already runs there.

All three point at the same model. Which one you pick is mostly about where your billing and infrastructure already sit.

Where Claude Fable 5 shines

Knowing what the model is built for saves you money and frustration. Anthropic framed this release around a handful of strengths, and they're worth matching to your actual work.

• Coding and software engineering. This is the headline use. Big refactors, multi-file changes, the kind of engineering work that takes real steps.
• Long context across very large inputs. We're talking millions of tokens. If you're working across a huge codebase or a stack of long documents at once, this is the standout. Holding focus across that much input is a different skill than just accepting it.
• Vision. It can read figures and rebuild a web app from a screenshot. Handy if your inputs are visual rather than plain text.
• Finance and analysis. Deep analytical work where the reasoning has to hold together over a long chain.

If your task fits one of those buckets, especially the long context and coding ones, Fable 5 is doing what it was made to do. For a quick one-off question or a simple script, you might not feel the difference, and the higher price won't be worth it. We get into who benefits in Claude Fable 5 for content creators if that's your angle.

When it hands off to Opus 4.8

This part surprises people, so it's good to know going in. Fable 5 ships with safety classifiers that hand a few request types back to Opus 4.8 instead of answering them directly. The areas are cybersecurity, biology and chemistry, and distillation, which means training other models on Fable's outputs.

In practice this triggers in under 5% of sessions. For most everyday work you'll never notice it. But if you're working in one of those topic areas and a response feels different, that's likely why: it came from Opus 4.8, not Fable 5. It's not an error on your end, it's the safety design doing its job.

Frequently asked questions

How do I use Claude Fable 5?

Two main paths. In the Claude apps or on claude.com, start a chat and select Claude Fable 5 in the model picker, which is included free on eligible plans from June 9 to June 22, 2026. For building, call the API and set the model to 'claude-fable-5'. It's also available through consumption based Enterprise and AWS Bedrock.

What is the model id for Claude Fable 5?

The model id is claude-fable-5. Pass that as the model value in your API request and your calls route to Fable 5.

Do I need to pay to use Claude Fable 5?

It depends on how you access it. On Pro, Max, Team, and seat based Enterprise plans it's free from June 9 to June 22, 2026, after which those plans need usage credits. On the API it's pay per token at $10 per million input tokens and $50 per million output tokens. That free window is time limited and could change.

What is Claude Fable 5 best at?

Coding and software engineering, long context work across very large inputs of millions of tokens, vision tasks like reading figures or rebuilding a web app from a screenshot, and finance and analysis. A small set of requests in cybersecurity, biology and chemistry, and distillation get handed back to Opus 4.8.