Anthropic Launches Claude Fable 5.1 & Mythos 5.1: Features & Pricing

Introduction

If Mission: Impossible had an Artificial Intelligence (AI) model lineup, Claude Fable 5.1 would be the field agent sent on difficult missions while Claude Mythos 5.1 would be the same agent carrying clearance for classified assignments.

That is surprisingly close to Anthropic's latest model strategy.

On September 1, 2026, Anthropic introduced Claude Fable 5.1 and Claude Mythos 5.1, calling them its most advanced models for coding and knowledge work. The two share the same underlying model but use different safeguards for advanced cybersecurity and life sciences work.

The launch also tackles another increasingly important AI problem: cost. While Anthropic has kept Fable's standard input and output prices unchanged, it has substantially reduced the price of repeatedly reading cached information.

So, is Claude Fable 5.1 merely Fable 5 with a decimal-point upgrade?

Not quite. The more meaningful changes appear in what the model can accomplish over long sessions and how much those sessions can cost.
 

Claude Fable 5.1 And Claude Mythos 5.1: Same Model, Different Clearance

Before comparing their features, there is an important distinction to make.

Claude Fable 5.1 and Claude Mythos 5.1 are the same underlying model with different levels of safeguards. Fable 5.1 is generally available while Mythos 5.1 is offered through Anthropic's trusted-access programs.

Their availability reflects that difference:
 

• Claude Fable 5.1: Available to paid Claude users and developers who need Anthropic's frontier model for coding, research, and enterprise work.

• Cloud availability: Developers can use Fable 5.1 through Amazon Web Services (AWS), Google Cloud, Microsoft Foundry, and Anthropic's own platform.

• AWS access: AWS confirmed that Fable 5.1 is available through Amazon Bedrock and Claude Platform on AWS.

• Claude Mythos 5.1: Limited to vetted individuals and organizations that require more permissive safeguards.

• Specialized use: Mythos focuses particularly on advanced cybersecurity and life sciences research.

That makes the Fable-versus-Mythos question less about which model is smarter and more about what level of access the work requires.
 

Anthropic AI Model Updates: What Claude Fable 5.1 Does Better

Claude Fable 5.1 keeps some familiar specifications. According to Anthropic's official model documentation, it supports a 1 million-token context window, up to 128,000 output tokens, and adaptive thinking that remains always on.


The bigger Anthropic AI model updates concern how Fable 5.1 performs across longer and more complicated tasks.
 

• Long-running coding: It can handle features spanning entire codebases, conduct code reviews, work on performance improvements, and continue autonomous sessions for hours.

• Better verification: The model can write tests to verify whether the code it generated actually works.

• Root-cause analysis: Anthropic says Fable 5.1 is less likely to take shortcuts that merely hide a problem and more likely to identify its underlying cause.

• Agentic workflows: It can plan work, use several tools, recover when a step fails, and continue operating with limited supervision.

• Knowledge work: Fable 5.1 can carry research from an initial question through analysis and into a finished deliverable.

• Vision capabilities: It can understand diagrams, charts, and tables inside documents and PDFs, which expands its usefulness for finance, legal, analytics, and other document-heavy workflows.

There is early customer evidence behind those claims too. Josh Boyer, Distinguished Engineer at Red Hat, said Fable 5.1 "correctly identified the root cause of every broken build we tested."

Developers migrating from Fable 5 should also note three breaking changes: forced tool use returns an error, earlier models cannot read Fable 5.1 thinking blocks, and modifying previous turns can invalidate those thinking blocks.
 

Claude Fable 5.1 Pricing: What Does It Actually Cost?

This may be the most commercially significant part of the launch.

The headline Claude Fable 5.1 pricing has not changed from Fable 5. Standard input still costs $10 per million tokens while output costs $50 per million tokens.

The major change is prompt caching

According to Anthropic's current Claude pricing documentation:
 

Pricing Element	Claude Fable 5.1	Claude Fable 5
Input	$10 / 1M tokens	$10 / 1M tokens
Output	$50 / 1M tokens	$50 / 1M tokens
5-minute cache write	$12.50 / 1M tokens	$12.50 / 1M tokens
1-hour cache write	$20 / 1M tokens	$20 / 1M tokens
Cache read	$0.25 / 1M tokens	$1 / 1M tokens

Here is what changed financially:
 

• Cache reads fell from $1 to $0.25 per million tokens.

• That makes Claude Fable 5.1 cache reads 75% cheaper than Fable 5.

• Anthropic estimates that typical token-billed workloads could cost around 25% less.

• For highly agentic workloads, Anthropic estimates savings of up to approximately 45%.

• Batch API processing receives a 50% discount on standard input and output pricing.

 

Claude Fable 5.1 Vs Claude Mythos 5.1: What Are The Real Differences?

The differences between Claude Fable 5.1 and Claude Mythos 5.1 primarily concern availability and safeguards rather than core intelligence.

Both models share:
 

• The same underlying model

• A 1 million-token context window

• Up to 128,000 output tokens

• Adaptive thinking

• $10 per million input tokens

• $50 per million output tokens

The divide appears when sensitive capabilities enter the picture.

Fable 5.1 can now be used to identify software vulnerabilities, making defensive cybersecurity work less likely to trigger restrictions. However, Anthropic still redirects several dual-use activities, including penetration testing, exploit generation, and binary-based vulnerability scanning.

Anthropic says the revised controls mean Claude Code users can expect around 60% fewer cybersecurity safeguard interventions per session than under Fable 5's earlier safeguards.

The biology safeguards have also become more precise. Anthropic says they now activate 85% less often on benign elementary biology and medical requests. More advanced life sciences work can instead be accessed through controlled Mythos programs.

There is our Mission: Impossible analogy completed. Fable and Mythos are not two different agents. They are effectively the same agent operating with different security clearances.
 

Do Claude Fable 5.1 Benchmarks Justify The Upgrade?

Anthropic's benchmark results indicate sizeable gains in several areas.

The company's official Fable 5.1 benchmark comparison reports:
 

• Terminal-Bench-Science 0.1: Fable 5.1 scored 52.6%, compared with 24.7% for Fable 5 and 29% for Opus 5.

• Terminal-Bench 4.0: Fable 5.1 reached 55.8%, versus Fable 5's 42%.

• AutomationBench: Its score increased from 17.1% on Fable 5 to 31.4%.

• CursorBench 3.2.0: Fable 5.1 scored 73.4%, compared with 70.5% for Fable 5 and 70% for Opus 5.

The more memorable evidence comes from practical experiments.

Investment firm Millennium had experienced a software crash roughly once in every million runs that its team had reportedly been unable to explain for four to five years. Fable 5.1 reportedly disassembled an external vendor library, compared it with the core dump, and traced the crash to that library.

Anthropic also used Fable 5.1 in planetary research. The model helped train a neural network that generated a new elevation map covering around one-third of Venus. The new map reveals details at roughly two to three kilometers instead of 10 to 20 kilometers and improves elevation estimates by up to 25%.

Mythos 5.1 went deeper into biological research. Anthropic reported that its protein designs achieved binding affinities 10 times higher than the strongest relevant designs in Adaptyv Bio competitions across three targets. Its hit rate also approached 50% across 12 targets, compared with Anthropic's stated typical rate of 10% to 15%.

These figures come from Anthropic's own evaluations and early partners, so they should not be treated as guaranteed results for every workload. They do, however, make the 5.1 release look more substantial than a routine model refresh.
 

Safety And Enterprise Controls Come With The Upgrade

Greater capability also introduces stricter requirements.

Anthropic has designated Fable 5.1 and Mythos 5.1 as Covered Models, a category reserved for models whose capabilities warrant additional safeguards and data-handling requirements.

For enterprises, the important points include:
 

• Fable 5.1 requires 30-day data retention for safety monitoring by default.

• Anthropic is introducing Enterprise Frontier Safeguards (EFS) for eligible organizations.

• EFS allows customers to keep monitoring data inside cloud infrastructure they control.

• Human review can remain with the customer rather than Anthropic by default.

Anthropic says EFS was developed with more than 100 enterprise customers across industries including finance, healthcare, manufacturing, telecommunications, law, retail, and the public sector.

This matters because enterprise AI buying is no longer a benchmark competition alone.

The strongest model may look impressive on a leaderboard but businesses still need to ask what it costs to operate, where their data goes, what capabilities they can access, and which safeguards could affect their workflows.

Fable 5.1 attempts to improve several parts of that equation at once.
 

Conclusion

Anthropic launches Claude Fable 5.1 and Mythos 5.1 at a moment when simply claiming to have a smarter AI model is no longer enough.

Fable 5.1 keeps premium base pricing but makes long-running agentic work more economical by reducing cache-read pricing by 75%. Mythos 5.1 takes the same underlying intelligence into more sensitive scientific and cybersecurity territory while limiting access to vetted users.

The release therefore says something broader about where frontier AI is heading. Model intelligence still matters but cost per task, safeguards, privacy, and the ability to operate reliably for hours are becoming just as important.

For that Mission: Impossible agent, intelligence was never enough. The mission also needed the right clearance, tools, and budget. Claude's 5.1 generation is beginning to look much the same.