TL;DR
Anthropic released Claude Fable 5, described by the company as its most capable model made generally available. The launch also introduces a split between public Fable 5 and restricted Mythos 5, with safety classifiers routing flagged prompts to Claude Opus 4.8.
Anthropic released Claude Fable 5 on June 9, 2026, calling it the most capable Claude model it has made generally available, a launch that matters because it brings the company’s higher-tier Mythos-class system to public users with new safeguards for risky requests.
According to the source material, Fable 5 and Mythos 5 use the same underlying model. Fable 5 is the public version with safety classifiers active, while Mythos 5 is a restricted version with safeguards lifted in some areas for selected partners.
The public model is listed for API use as claude-fable-5. The restricted Mythos 5 version is being deployed first through Project Glasswing, Anthropic’s cyber-defense program with the U.S. government, and is expected to reach selected biology researchers later.
The key design choice is that Fable 5 does not simply refuse some flagged prompts. The source material says requests identified as risky in areas such as cybersecurity, biology, chemistry, or distillation are routed to Claude Opus 4.8, a weaker model. Anthropic says users are told when that routing happens. The classifiers are described as conservative, so some benign requests may be sent to Opus 4.8.
Fable & Mythos
Anthropic just shipped its most capable public model — and the story is how. One “Mythos-class” model, two names, and a safety net that hands risky queries to a weaker model instead of refusing them.
- The best coding model in the world they’ve tested — 91/100, near human-engineer range.
- Paradigm-shifting for power users on their hardest, long-horizon tasks.
- One-shots entire apps; owns a whole job end-to-end over multi-hour runs.
- Overpowered for everyone else — lower-adoption users struggled to find a use.
- Slow & token-hungry; ~2× Opus 4.8 cost, >3× Sonnet 4.6. Mixed for writing.
- Rewards a sharp brief, punishes a loose one — precision in, precision out.
Independent commentary, produced with AI assistance under human editorial oversight. The views are the author’s own and may change. This is analysis, not investment, financial, legal, or technical advice. Details of Claude Fable 5 and Mythos 5 — capabilities, safeguards, pricing, rollout, and figures — are drawn from Anthropic’s launch announcement and Every’s independent “Vibe Check,” both June 2026, and may change as the models and access terms evolve. Benchmarks and testimonials are as reported by their sources. Company and product names are referenced for analysis and imply no affiliation or endorsement.
A New Public Safety Model
The release is less about a single benchmark and more about how frontier AI systems may be shipped. Anthropic is making a more capable model broadly available while keeping part of its risk management inside the product path, rather than relying only on refusals or limiting the whole system to private access.
For developers and companies, that could change how advanced models are adopted. The source material frames Fable 5 as better suited to long-running, high-value jobs than ordinary chat use, including codebase migrations, multi-hour agent work, and tasks that benefit from detailed briefs and careful review.
The economics also matter. Fable 5 is reported at $10 per million input tokens and $50 per million output tokens, less than half the price of Mythos Preview but about twice the cost of Opus 4.8. That makes model selection a business decision as much as a capability choice.
Unlocking Data with Generative AI and RAG: Enhance generative AI systems by integrating internal data with large language models using RAG
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
From Opus To Mythos
Anthropic’s prior public model line placed Opus at the top of its generally available family. Fable 5 is described in the source material as the first public Mythos-class model, a tier above Opus that Anthropic had previously treated as too risky for broad release.
Early access reviewers at Every tested Fable 5 for a week and reported strong coding results. Every gave it a 91 out of 100 on its Senior Engineer benchmark, compared with 63 for Claude Opus 4.8 and 62 for GPT-5.5, according to the figures cited in the source material.
Other reported examples include a Stripe codebase-wide migration across a 50 million-line Ruby codebase that the source says took one day instead of two months by a team, faster drug-design work using Mythos 5, and strong vision-based performance on software reconstruction and game play. Those results are reported by Anthropic, Every, or named partners and have not been independently verified here.
“most capable model it has ever made generally available”
— Anthropic
“the best coding model in the world”
— Every
“a warp drive for power users”
— Every
The AI Cybersecurity Handbook
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Open Questions On Routing
It is not yet clear how often ordinary users will see prompts routed from Fable 5 to Opus 4.8, or how many legitimate security and life-science workflows will be affected by conservative classifiers. The source material says flagged requests are expected to be under 5%, but real-world rates may vary by user group and task type.
Independent verification is also limited. Several performance examples come from Anthropic, Every, or customer accounts cited in the launch material. Details about Mythos 5 access, safeguards lifted for partners, and model retention policies may change as Anthropic expands or revises the rollout.
Ocean Restoration Digital Systems: Blue Economy, AI Marine Monitoring, Blockchain, Biotechnology, Smart Fisheries Governance, Infrastructure
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Credits, Access And Testing
Fable 5 is available through June 22 to Pro, Max, Team, and Enterprise users under the launch terms described in the source material. After that, Anthropic plans to move users to usage credits and later standard access.
The next tests will be practical ones: whether developers can control cost on long tasks, whether the routing system avoids blocking legitimate work, and whether Anthropic expands Mythos 5 beyond Project Glasswing and selected research partners without changing the risk model.
Llama Guard in Practice: Safety Classification Pipelines for LLM Responses
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did Anthropic release?
Anthropic released Claude Fable 5, a public Mythos-class model, and described it as its most capable generally available Claude model.
Is Fable 5 the same as Mythos 5?
According to the source material, Fable 5 and Mythos 5 use the same underlying model. Fable 5 has public safeguards active, while Mythos 5 has some safeguards lifted for selected partners.
How does the safety system work?
Fable 5 uses classifiers to detect risky prompts in areas such as cybersecurity, biology, chemistry, and distillation. Flagged prompts are routed to Claude Opus 4.8 instead of being answered by the full Fable 5 model.
Who can use Mythos 5?
Mythos 5 is restricted to trusted partners, starting with Project Glasswing cyber-defenders. Selected biology researchers are expected to gain access later, according to the source material.
Why does this release matter?
The launch shows one way AI companies may bring stronger models to public users while limiting access to higher-risk capabilities. It also gives developers a more expensive but more powerful option for long-running coding and research tasks.
Source: Thorsten Meyer AI
