Claude Fable 5, Anthropic's newest flagship model, is now globally available after being offline for 18 days due to export control restrictions. The model launched June 9, 2026, as the first publicly available member of Anthropic's "Mythos-class" tier, but was suspended on June 12 after Amazon researchers discovered a technique that bypassed its cybersecurity safeguards. Access was restored worldwide on July 1 with an updated safety classifier that Anthropic says blocks the reported technique in over 99% of cases.

What Makes Claude Fable 5 Different From Earlier Claude Models?

Fable 5 represents a significant leap in capability compared to Anthropic's previous flagship, Claude Opus 4.8. On SWE-Bench Pro, a benchmark that tests real-world software engineering tasks across multiple files, Fable 5 scored 80.3%, compared to Opus 4.8's 69.2% and OpenAI's GPT-5.5 at 58.6%. The model also achieved a 95% score on SWE-Bench Verified and topped Cognition's FrontierCode evaluation, with Stripe publicly endorsing the model for compressing "months of engineering into days".

The underlying architecture is the same as Claude Mythos 5, a restricted sibling available only to vetted cybersecurity and biosecurity researchers through Anthropic's Project Glasswing program. The key difference is that Fable 5 ships with safety classifiers that monitor requests for potential misuse in three areas: cybersecurity exploitation, biology and chemistry dual-use research, and attempts to distill the model's outputs for training competing systems.

When a request triggers one of these classifiers, the model doesn't answer it. Instead, Anthropic automatically routes the request to Claude Opus 4.8, its more conservative predecessor. According to Anthropic, these safeguard triggers fire in less than 5% of sessions on average, and declined requests don't incur charges.

How to Access Claude Fable 5 and What It Costs?

• API Access: Available through the standard Claude API using the model ID "claude-fable-5," as well as Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry at $10 per million input tokens and $50 per million output tokens.
• Claude.ai Plans: Accessible on Claude Pro, Max, Team, and Enterprise subscription tiers. Between June 9 and June 22, access was included at no extra cost; since then it draws on usage credits.
• Claude Code: Available in Claude Code, Anthropic's AI-powered coding environment, where extended thinking is always enabled and cannot be disabled.

At roughly double the price of Opus 4.8, Fable 5 is positioned for genuinely difficult tasks rather than routine work. Anthropic's own documentation recommends Opus 4.8 as the default for most enterprise work, reserving Fable 5 for long multi-step agentic tasks, research-grade analysis, complex codebases, and vision-heavy workloads where a weaker model's failure would cost more than the price difference.

The model ships with a 1 million token context window, meaning it can process roughly 100,000 words in a single request without needing retrieval scaffolding. This is particularly valuable for analyzing entire codebases or document sets in one go. Maximum output is capped at 128,000 tokens per request, and the knowledge cutoff is January 2026, the most current of any Claude model.

Why Was Claude Fable 5 Suspended, and What Was the Fix?

The suspension began on June 12, three days after launch, when Amazon researchers reported a method that bypassed Fable 5's cybersecurity safeguards. In at least one documented case, the model produced code demonstrating how the relevant vulnerability could be exploited. Unable to verify user nationality in real time, Anthropic suspended both Fable 5 and Mythos 5 globally while the US government applied export controls to the newest models.

The timing raised questions about Anthropic's messaging. Just days before the Fable 5 launch, the company had urged the AI industry to agree on a "coordinated brake pedal" for frontier AI development, warning that systems could soon reach recursive self-improvement without human oversight. Then it shipped its most capable public model anyway. Anthropic's answer to that tension was the safety architecture wrapped around Fable 5 rather than a delay to the release schedule.

The company's response to the Amazon discovery was swift. Anthropic retrained its safety classifier and restored access on July 1 across the Claude Platform, claude.ai, Claude Code, and Claude Cowork, with cloud marketplace reinstatement on AWS, Google Cloud, and Microsoft Foundry following. The updated classifier blocks the reported technique in over 99% of cases, with blocked requests falling back to Claude Opus 4.8.

What Does This Mean for Open-Source Development and AI Coding Tools?

The broader context of Fable 5's launch includes growing concerns about AI coding agents like Claude Code and Cursor raising code complexity and potentially deterring newcomers from open-source projects. A recent analysis of 1,888 GitHub projects tracked adoption of AI coding agents and found that newcomer participation remained steady despite increased code complexity.

The study found that while Python's cognitive complexity metrics increased by about 11% and cyclomatic complexity across languages rose 3-4%, newcomer inflow and retention rates were unchanged. This suggests that open-source culture is resilient enough to integrate AI tools without sidelining human contributors. Rather than automating newcomers out of projects, AI coding agents may be pushing entry-level developers to tackle more complex challenges earlier, potentially accelerating their learning curve.

For developers using Fable 5 in production, the key takeaway is understanding how the safety classifiers work. If your workload involves security research or the life sciences, Anthropic recommends testing against these classifiers before committing to the model. For everyone else, the classifiers will rarely trigger, and the 1 million token context window is the genuinely underrated feature that sets Fable 5 apart from its predecessors.