Anthropic released Claude Fable 5, a next-generation "Mythos-class" AI model, on Tuesday. The model is designed to restrict dangerous capabilities in areas such as cybersecurity and biological research after CEO Dario Amodei warned about risks last month.
The model gives users access to Anthropic's more powerful Mythos model, which the company had previously deemed too risky for public release last month. However, when users ask about sensitive topics, such as bioweapons or software exploitation, Fable 5 redirects them to the older Claude Opus 4.8 model.
"We maintain that Anthropic is the leading pure-play AI lab, combining best-in-class model intelligence with its cutting-edge, benchmark-leading Claude Fable 5 frontier model released June 9, 2026; with clear commercial traction and momentum in its enterprise offerings," BMO analyst Brian Pitz wrote in a note earlier today.
Pitz noted, "Anthropic's strengths are particularly evident in coding, agents, and enterprise, where Claude has emerged as a leading model powering tools such as Claude Code and Cowork, both of which have scaled rapidly. This reinforces the company's advantage in translating model intelligence beyond benchmark performance into viable, real-world applications—what we view as the next key battleground in AI."
The release of Claude Fable 5 prompted Pitz's team to declare, "While it is too early to crown a winner among foundation models, we see Anthropic and OpenAI as the leading pure-play AI labs today."
The Verge's Tom Warren reported that Claude Fable 5 has already raised security concerns within Microsoft, prompting the tech giant to limit internal employee access to the model due to Anthropic's data-retention requirements.
Warren said that Claude Fable 5 has been rolled out to GitHub Copilot and Foundry customers but is not available in the internal GitHub Copilot model picker used by Microsoft employees. Other Claude models remain available internally because they operate under zero data retention rules.
He said the issue centers around Anthropic's safety architecture. Claude Fable 5 requires Anthropic to retain prompts and outputs for 30 days to operate new safety classifiers, while some flagged content can be stored for up to two years if it violates usage policies. These rules could potentially create risks for confidential information.
Pitz published the current AI leaderboard overview with Anthropic's models on top (but at the time of the note, Claude Fable 5 was not included):
Western AI Models Comparison
BMO analysts see the release of new advanced models driving AI revenue to $1.8 trillion by 2032. That would mean the market has expanded at an average annual growth rate of 48% since ChatGPT launched in 2022.
Token prices have declined over the last six days.
"Adoption is becoming less about what frontier models can do and more about the price... the recent drop in the token index may reflect some of this shift toward cheaper models," Citadel analysts noted (read).
Prices per million tokens for Western models vs. Chinese models
Tokenmaxxing.
Average cost per task.
What X users have been creating with Claude Fable 5: