Claude Fable 5.1 Explained: Which Plan Gets It and What It Costs

Anthropic released a new top model on September 1, 2026, and for almost everyone paying for Claude the honest answer is that nothing about your plan changed. Claude Fable 5.1 and Claude Mythos 5.1 are, in Anthropic's words, "the world's most advanced models for coding and knowledge work." The consumer questions underneath that are the same three as always: do you get it, does it cost more, and does your Claude feel different today.

Key Takeaways

• Free Claude gets nothing. The Fable row is marked not included on the Free plan on claude.com/pricing, and 5.1 does not change it. Free Claude still defaults to Sonnet 5.
• Claude Pro at $20 still pays extra for it. Fable bills against pay-as-you-go usage credits on Pro. That has been true since the inclusion promotion ended on July 19, 2026, and the point release did not reset it.
• Max is where it is included, up to 50% of your weekly limits on both Max 5x and Max 20x.
• The price cut is for developers, not subscribers. Cache reads fell 75%, to $0.25 per million tokens. Input and output are unchanged at $10 and $50 per million. Anthropic estimates about 25% less for typical workloads and "up to approximately 45%" for heavily agentic work.
• It is the first Claude that marks its own writing. Anthropic's help centre names Fable 5.1 and Mythos 5.1 as the models "currently supported" by its invisible text watermark.
• Claude may switch models on you mid-conversation, and the setting to stop that is in Settings > Capabilities.
• Mythos 5.1 is not a product you can buy. Same model, fewer guardrails, restricted to vetted cyberdefenders and life scientists at "a set of US organizations."
• Safeguards got less trigger-happy. Anthropic says its newest cybersecurity safeguards "block 60% fewer false positives than before."

Which Claude Plans Get Fable 5.1

Exactly the plans that got Fable 5, on exactly the same terms. This is a point release, not a repackaging, and the plan comparison matrix on Claude's pricing page did not move with it.

Decoded from the plan comparison matrix on claude.com/pricing on September 8, 2026, and from the availability section of Anthropic's launch post. Anthropic says Fable 5.1 "is available today on all platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure."

The one number that did change for a paying subscriber is indirect. Because Fable on Pro bills by usage credits, and usage credits are billed by token, the 75% cut to cache reads reaches Pro users too. Anthropic publishes no consumer-facing figure for it, so treat "about 25% cheaper for typical workloads" as its estimate for token-billed work generally, not a promise about your credit balance.

What Actually Got Better

Coding, long-running tasks and a cost curve that lets you spend less for the same result. Anthropic's framing is that Fable 5.1 "sets a new standard for coding, knowledge work, and long-running problem-solving tasks," and that "when set to Low or Medium effort, Fable 5.1 achieves results similar to or better than Fable 5's at a much lower cost."

That second half is the part with a consumer edge. There is an effort setting, and where you meet Claude decides its default:

So if you use Fable 5.1 in an ordinary chat, you are on Medium by default, which is the setting Anthropic says matches or beats the old model at lower cost. You do not have to do anything to get that.

On long documents it starts where the other paid models are. Anthropic's help centre states that "Claude Fable 5.1, Opus 5, and Sonnet 5 support a 1M token context window on all paid plans when chatting with Claude," so 5.1 does not buy you more room than Pro's included Opus 5 already gives you. Note that the plan comparison table on the pricing page shows a flat 200k in its context-window row for every plan; that is the baseline for models outside that named list, and the help centre is the surface that answers the question for these three.

Three other changes Anthropic published, in plain terms:

• Fewer wrong refusals in cybersecurity. The newest safeguards "block 60% fewer false positives than before," and Fable 5.1 "can now be used to discover software vulnerabilities -- though not to develop exploits for them."
• A biology access programme is coming, built "in partnership with the US government," to let vetted scientists reach Mythos 5.1's biology capabilities. Anthropic says it expects "to open enrollment for scientists soon" and publishes no date.
• Enterprise Frontier Safeguards, a data-retention system that keeps customer data in the customer's own cloud, arrives "in phases, beginning later this fall." This one is for enterprise buyers, not for you.

Why Claude Might Switch Models Mid-Conversation

Because Fable 5.1 runs a safety check on every request, and a flagged request gets re-run on a different model. This has been true since Fable 5 and it carries over unchanged, but it is the single most confusing thing about using Fable and it is worth knowing before it happens to you.

Anthropic's help article on the behaviour is specific about which requests fall back:

Two details in that article catch people out. First, the checks read everything the model reads, "not just your latest message -- including memory, content from connectors, web search results, and files, so a block can be triggered by content you didn't type." Second, after a switch the model picker stays on Opus for the rest of the conversation. You can switch back manually, but Anthropic warns that if the original request is still in the chat it may simply fall back again, and suggests editing your earlier message first.

To turn the behaviour off: Settings > Capabilities, then toggle Switch models when a message is flagged off. With it off, a blocked request pauses the conversation instead of quietly moving you. Anthropic says automatic switching works the same way on Claude on the web, mobile, desktop, Cowork, Claude Code, Claude Design, Claude for Microsoft 365 and Claude Tag.

On billing, the rule is fair and worth knowing if you are on usage credits: if a request is blocked on input, "you're charged only at Opus rates." If it is blocked midstream, the tokens already streamed are charged at Fable rates and the rest at Opus rates.

Fable 5.1 Is the First Claude That Watermarks Its Writing

This is the genuinely new thing for an ordinary reader, and it arrived on the same day. Anthropic's help centre article on marking AI-generated content, revised September 1, 2026, states: "Models currently supported include Fable 5.1 and Mythos 5.1." Everything older is covered by an EU transition period Anthropic is still working through.

What that means in practice: text written by Fable 5.1 carries an invisible watermark in its word choices. You cannot see it, it adds nothing to the text, it costs nothing, and Anthropic says it "carries no identifying information and can't be traced to a specific person, organization, or chat." Checking it requires Anthropic's detection API, which went into private preview on September 1, 2026 for regulators, media, researchers, educational organisations and enterprises with an EU compliance obligation. Our full explainer on Claude's text watermark covers what it can and cannot prove.

The practical consequence: if you are on Pro and you want unwatermarked Claude output, the default Sonnet 5 is currently not on Anthropic's marking list and Fable 5.1 is. Anthropic has said it intends to bring the rest of the family in, so treat that as today's state rather than a lasting difference.

Should You Change Anything?

No, if you are on free Claude. Fable has never been on the free plan and 5.1 does not change that. Nothing about your Claude moved on September 1. If you were considering paying, the reasons are the same as they were last month, and which AI chatbot is worth paying for walks through them.

No, if you are on Pro and were not already buying usage credits. The upgrade path here has not got cheaper in a way you will notice. Pro's included ceiling is still Opus 5, which Anthropic calls the strongest model included with Pro, and Opus 5 is genuinely good.

Maybe, if you are on Pro and already spend on credits. Cache reads getting 75% cheaper is a real saving on long, context-heavy sessions, and Fable 5.1 at Medium effort is Anthropic's claim to match the old model for less. If your credit spend is a monthly line item, this is the month to look at it.

Yes, in one small way, whatever plan you are on: know that Settings > Capabilities controls whether Claude silently swaps models on you. That surprises people far more than a point release does.

The bottom line: Fable 5.1 is a developer-and-enterprise release wearing a consumer name. The plan matrix is unchanged, the free tier is unchanged, and the price cut lands on tokens rather than subscriptions. The two things worth carrying away are that it is the first Claude marking its own text, and that Claude will move you to a different model mid-chat unless you tell it not to.

Sources

• Anthropic, Introducing Claude Fable 5.1 and Claude Mythos 5.1, September 2026
• Anthropic, Release notes, help centre, entry dated September 1, 2026: "Claude Fable 5.1 and Claude Mythos 5.1 launch"
• Anthropic, Why Claude switched models in your conversation with Fable 5 or Fable 5.1, last revised September 1, 2026
• Anthropic, How Claude marks AI-generated content, last revised September 1, 2026
• Anthropic, How large is the context window on paid Claude plans?, last revised September 1, 2026
• Anthropic, claude.com/pricing plan comparison matrix, decoded September 8, 2026