Claude Fable 5: Available on Google Cloud

Overview

Claude Fable 5, Anthropic's latest frontier model, is now generally available on Google Cloud's Agent Platform as of June 9, 2026. Google positions the release as part of an ongoing effort to bring the newest models to its platform with safeguards built in for general business use. The model targets demanding work such as advanced software development, long-running autonomous agents, and detailed document analysis. It joins Claude Opus 4.8 and Claude Sonnet 4.6, which are already available through the same platform.

Key Takeaways

Claude Fable 5 reached general availability on Google Cloud's Agent Platform on June 9, 2026, meaning business customers can build with it in production rather than a limited preview.
Anthropic built the model for complex, multi-step reasoning, with stated strengths in software development, long-horizon agents, and multimodal document analysis.
Google ships the model with what it calls strong safeguards designed to keep it safe for general use, including safety classifiers that block high-risk requests.
Fable 5 sits alongside two other Anthropic models on the platform, Claude Opus 4.8 and Claude Sonnet 4.6, giving teams a range of price and capability options.
Pricing runs $10 per million input tokens and $50 per million output tokens, less than half the cost of the earlier Claude Mythos Preview model.
Anthropic reports state-of-the-art benchmark scores and customer results, including Stripe completing a 50-million-line codebase migration in a single day.

Stats & Key Facts

#$10 per million input tokens, the published input price for Claude Fable 5
#$50 per million output tokens, the published output price, less than half the cost of Claude Mythos Preview
#50-million-line codebase migration completed in one day by Stripe, a task estimated at two months for a team
#1,000-plus hours of external red-teaming found no universal jailbreak against the safeguards
#Over 95 percent of Fable sessions involve no fallback to the Opus 4.8 model
#30-day data retention requirement applies to all traffic, with data deleted afterward and used only for safety

Claude Fable 5 Reaches General Availability on Google Cloud

The headline of the announcement is a status change from preview to production.

Google Cloud confirmed that Claude Fable 5, Anthropic's latest frontier model, is now generally available on its Agent Platform. General availability is the stage where a product is considered ready for production use, not just testing, so business customers can build live applications on the model with standard support.

Google framed the launch as the latest proof point of an ongoing commitment to bring the industry's newest models directly to its Agent Platform. The release date was June 9, 2026, the same day Anthropic announced the model more broadly.

Built for Complex Reasoning, Coding, and Long-Running Agents

Anthropic designed Fable 5 for the hardest categories of work rather than quick lookups.

›Advanced software development, including large codebase changes and engineering tasks
›Long-horizon agents that run multi-step work over extended periods
›Deep multimodal document analysis that combines text and images
›Autonomous knowledge work that runs in the background without constant supervision

The common thread across these uses is sustained, multi-step reasoning. Anthropic positions Fable 5 as a model for ambitious, long-running projects rather than short single-turn requests.

Safety Safeguards and a 30-Day Data Rule

Google stressed that the model ships with protections aimed at general business use.

›Classifiers block cybersecurity requests tied to exploitation and offensive cyber tasks
›Classifiers block dual-use biology and chemistry requests, falling back to the Opus 4.8 model when triggered
›A separate system blocks attempts to extract the model's capabilities for competing systems

Anthropic reports that more than 1,000 hours of external red-teaming found no universal jailbreak, and that over 95 percent of Fable sessions run without any fallback to Opus 4.8. Fable 5 also carries a 30-day data retention requirement, where traffic is kept for safety purposes only and deleted after that window. Organizations that require zero data retention cannot use the model under that configuration.

How Fable 5 Fits Alongside Opus 4.8 and Sonnet 4.6

Fable 5 is not the only Anthropic option on the platform.

›Claude Fable 5, the most capable and most expensive option, for the most demanding work
›Claude Opus 4.8, a highly capable model at a lower price point
›Claude Sonnet 4.6, positioned for a balance of speed and intelligence

Having three models on one platform lets teams match the model to the task and budget. Routine work can run on Sonnet or Opus, while the heaviest reasoning and agent jobs go to Fable 5. The safety design reflects this tiering as well, since blocked Fable 5 requests fall back to Opus 4.8.

Pricing and Reported Performance

Anthropic published both a price and a set of customer and benchmark results.

›Input priced at $10 per million tokens
›Output priced at $50 per million tokens
›Less than half the per-token cost of the earlier Claude Mythos Preview

On results, Anthropic cites state-of-the-art scores across nearly all tested benchmarks, including the highest marks on Cognition's FrontierCode evaluation and Hebbia's Finance Benchmark. Customer examples include Stripe, which reported migrating a 50-million-line codebase in a single day, a task it estimated would have taken a team two months. These figures come from Anthropic's own announcement rather than independent testing.

What This Means for Non-Technical Teams

Here is the plain-language read on why this matters.

For a business already on Google Cloud, general availability means Fable 5 is now a supported building block for products and internal tools, not an experiment. Teams that need long-running automation, heavy document review, or large coding projects gain access to a model built for that scale.

Cost discipline still matters. At $50 per million output tokens, Fable 5 is the priciest of the three Anthropic models on the platform, so many teams will reserve it for the hardest tasks and run everyday work on Opus 4.8 or Sonnet 4.6. The 30-day data retention rule is also worth checking against any internal privacy policy before adoption.

Frequently Asked Questions

What is Claude Fable 5?

It is Anthropic's latest frontier model, built for complex multi-step reasoning. Anthropic targets it at advanced software development, long-running agents, and detailed multimodal document analysis.

What does generally available on Google Cloud mean?

General availability means the model has moved past preview and is ready for production use on Google Cloud's Agent Platform. Business customers can build and run live applications on it with standard support.

Which other Anthropic models are on Google Cloud's Agent Platform?

Claude Opus 4.8 and Claude Sonnet 4.6 are both available alongside Fable 5. This gives teams a range of capability and price options to match the task.

How much does Claude Fable 5 cost?

Anthropic lists pricing at $10 per million input tokens and $50 per million output tokens. That is less than half the per-token cost of the earlier Claude Mythos Preview model.

Is Claude Fable 5 safe for general business use?

Google describes the model as shipping with strong safeguards for general use, and Anthropic uses classifiers that block high-risk cybersecurity and biology requests. External red-teaming of more than 1,000 hours reported no universal jailbreak, though a 30-day data retention requirement applies.

Claude Fable 5's general availability on Google Cloud gives business teams a production-ready frontier model for their most demanding reasoning, coding, and agent workloads, sitting alongside Opus 4.8 and Sonnet 4.6 as a higher-cost, higher-capability option.