TL;DR
Anthropic released Claude Fable 5, its first broadly available Mythos-class model, with safeguards that route flagged requests to Opus 4.8. Claude Mythos 5 uses the same underlying model but remains limited to trusted partners, while early reviewers at Every called Fable 5 the best coding model they have tested.
Anthropic released Claude Fable 5 on June 9, 2026, making its first Mythos-class AI model broadly available while keeping a less restricted version, Claude Mythos 5, limited to trusted partners. The launch matters because Anthropic says the same underlying model had previously been judged too risky for broad release, and the company is now testing a safety design that routes flagged requests to Claude Opus 4.8 instead of simply refusing them.
Anthropic describes Fable 5 as the most capable Claude model it has made generally available. The company says Fable 5 and Mythos 5 are the same underlying model, with safeguards separating the public version from the partner-only version. Fable 5 is available through the API as claude-fable-5, priced at $10 per million input tokens and $50 per million output tokens, according to the launch materials.
The safety system watches for risky requests in areas including cybersecurity, biology, chemistry and model distillation. Anthropic says most requests are answered by Fable 5, while flagged requests are routed to Claude Opus 4.8. The company says users are told when routing happens, and the source material says the system is tuned conservatively enough that some benign requests may be caught.
Performance claims are still source-reported. Anthropic cited a Stripe codebase migration across a 50-million-line Ruby codebase that it said was completed in a day instead of two months by a team. Every, which had a week of early access, gave Fable 5 a 91 out of 100 on its Senior Engineer benchmark, compared with 63 for Opus 4.8 and 62 for GPT-5.5, and described the model as strongest on longer, harder coding tasks.
Guardrails Shape Model Access
The release is a test of whether frontier AI companies can ship more capable systems to broad audiences by using routing, pricing and partner restrictions instead of keeping the strongest models fully private. For developers and companies, the immediate issue is not only whether Fable 5 is more capable, but when its cost, latency and safety routing make it the right tool.
The launch also gives builders a clearer split between everyday model use and long-running agent work. Every’s review said Fable 5 performs best when given a clear, substantial task, such as building an app or completing a multi-hour engineering job. That framing points to a shift in workflow: users may spend more time defining the job, running agents in parallel and reviewing output, rather than treating the model as a fast chat assistant.
The safety design has business consequences. Security researchers, life-science teams and regulated companies may find that legitimate work is sometimes routed to a weaker model. That could reduce risk, but it may also affect reliability for the very users most likely to need frontier capabilities.
AI VoiceWriter – Smart Dictation & AI Writing Assistant for Windows & Mac | USB Dongle & Mobile App for Voice Input, Proofreading, Rewriting & Multilingual Support
- Hands-Free Voice Typing: Speech-to-text for Windows & Mac
- AI Writing Assistant: Proofreading, rephrasing, formatting help
- Universal App Compatibility: Works with Word, Google Docs, and more
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How Mythos Reached Public Release
Before this launch, Anthropic kept Mythos-class capability behind limited-access programs, including Project Glasswing, a cyber-defense effort involving trusted partners and the U.S. government. Mythos 5 remains restricted to that kind of access, with select biology researchers expected to be added later, according to the source material.
The naming reflects the access split. Anthropic says Fable and Mythos are related terms, both tied to what is told, but the practical difference is not the core model. It is what the model is allowed to do for different users.
The release follows a period in which frontier model labs have faced pressure to show both stronger capabilities and more convincing controls. Anthropic’s approach here is to make the stronger model public in a constrained form, while preserving a less restricted version for vetted use cases.
“the most capable model it has ever made generally available”
— Anthropic launch materials
Benchmarks Need Outside Testing
It is not yet clear how Fable 5 will perform across a wider set of users, codebases and production environments. The strongest performance figures currently come from Anthropic and Every, and independent replication will matter for companies deciding whether to move high-value work to the model.
The routing system also needs real-world testing. Anthropic says fewer than 5% of requests are expected to be flagged, but the practical effect on security research, biology workflows and compliance reviews remains developing. The timeline for broader Mythos 5 access is also not confirmed.
Rollout Moves to Paid Use
Anthropic is offering Fable 5 free on Pro, Max, Team and Enterprise plans through June 22, 2026, according to the source material. After that, access is expected to move to usage credits before standard pricing applies more broadly.
Developers can begin testing the API immediately, while trusted partners continue using Mythos 5 through restricted programs. The next signals to watch are external benchmarks, production case studies, changes to classifier routing and any expansion of Mythos 5 access beyond Project Glasswing.
Key Questions
Is Claude Fable 5 the same model as Claude Mythos 5?
Anthropic says they use the same underlying model. Fable 5 is the safeguarded public version, while Mythos 5 has fewer restrictions in some areas and is limited to trusted partners.
Does Fable 5 refuse risky prompts?
Anthropic says flagged requests are routed to Claude Opus 4.8 rather than handled by Fable 5. The company says users are told when that routing occurs.
How much does Claude Fable 5 cost?
The API price listed in the source material is $10 per million input tokens and $50 per million output tokens. Access is free on some paid plans through June 22, 2026.
Who can use Claude Mythos 5?
Mythos 5 is restricted to trusted partners, starting with Project Glasswing cyber-defense users. Select biology researchers are expected to receive access later.
What remains unverified?
The main open questions are how Fable 5 performs outside early testing, how often benign work is routed to Opus 4.8 and when Anthropic will broaden access to Mythos 5.
Source: Thorsten Meyer AI
