Anthropic released Claude Fable 5 on June 9, 2026, opening its most powerful AI to date for general business use. This is a significant moment: for the first time, the capabilities behind Claude Mythos — previously restricted to vetted government-adjacent cybersecurity partners through Project Glasswing — are available to any team using the Claude API or major cloud platforms.

Here is what the launch means in practice.

What Is Claude Fable 5?

Claude Fable 5 is Anthropic’s most capable widely-released model, designed for demanding reasoning and long-horizon agentic work. It shares the same underlying capabilities as Claude Mythos 5, a restricted version available only to approved Project Glasswing customers.

The distinction between the two is access and safety controls. Claude Mythos 5 has no safety classifiers and is reserved for vetted partners. Claude Fable 5 includes classifiers that can decline certain categories of requests — specifically in areas like cybersecurity, biology and chemistry, and model distillation. When a request triggers a classifier, the model falls back to Claude Opus 4.8 to handle it. Across typical business workflows, this affects fewer than 5% of sessions.

Both models share the same specifications:

• Context window: 1 million tokens, with up to 128,000 output tokens per request
• Pricing: $10 per million input tokens and $50 per million output tokens
• Thinking mode: Adaptive thinking is always on — there is no option to disable it

The 1 million token context window is the spec that will matter most to teams running large documents, extended multi-step workflows, or agents that need to hold significant context in memory across a long session.

Where You Can Access It

Claude Fable 5 is generally available on:

• The Claude API
• Claude Platform on AWS
• Amazon Bedrock
• Vertex AI
• Microsoft Foundry

This broad availability means most enterprise teams can access it through infrastructure they already use.

Claude Mythos 5 remains restricted to Project Glasswing partners. For teams outside that program, Claude Fable 5 delivers the same capabilities.

Pricing and Plan Access

Fable 5 was included on Pro, Max, Team, and Enterprise plans at no additional cost through June 22, 2026. After that date, usage credits may be required while Anthropic scales up capacity.

On a per-token basis, the pricing is double Claude Opus 4.8 but less than half the cost of Claude Mythos Preview. For teams running production workloads that demand higher-quality output or longer context, the improved capability-to-cost ratio is real.

One important data retention note: both Fable 5 and Mythos 5 carry a 30-day data retention policy and are not available under zero data retention agreements. Teams with strict data handling or compliance requirements need to account for this before migrating production workloads.

How the Safety Classifiers Work

When Claude Fable 5 declines a request, it returns a successful HTTP 200 response rather than an error. The response includes a stop reason of “refusal” and identifies which classifier triggered it.

For developers building on top of Fable 5, Anthropic has provided three fallback paths:

• Server-side fallback: Pass a fallbacks parameter to the API, which handles the retry automatically. Currently in beta.
• Client-side fallback: Use SDK middleware available for TypeScript, Python, Go, Java, and C#.
• Manual retry: Build your own retry logic in any language on any platform.

If a request is refused before any output is generated, it is not billed. If you retry on another model, Anthropic provides a fallback credit that offsets the prompt-cache cost of switching so you do not pay twice.

What This Means for Business

The availability of Fable 5 changes the calculus for enterprise AI workflows in two concrete ways.

First, the 1 million token context window removes a genuine constraint. Many business use cases — processing large documents, running extended agent sessions, automating research-heavy workflows — previously required complex engineering workarounds: chunking, summarisation pipelines, retrieval systems. Fable 5 can hold far more in context without those patches.

Second, the model’s performance across software engineering, knowledge work, and scientific reasoning means teams building internal tools, automating document-heavy processes, or running agentic workflows have a meaningfully better foundation than they did last week.

The step up in pricing from Opus 4.8 is real. But for tasks where output quality matters or where a less capable model requires more retries and human review to produce acceptable results, the economics can still favour Fable 5.

The businesses that will get the most from this launch are the ones that already have structured agentic workflows in place. If you have built agents on Opus 4.8 and have a clear picture of where quality has been the bottleneck, this is a straightforward upgrade decision. If you are still in the early stages of deploying AI agents in your business, the model capability is less important than having the right processes to put it to work.

If your team is evaluating whether to migrate production workflows to Fable 5, or building agentic systems that can take advantage of the extended context window, our Omni Advisory team can help you assess the architecture and cost trade-offs before you commit.