Anthropic has launched Claude Fable 5, a new artificial intelligence model that brings many of the advanced capabilities of its Mythos-class systems to regular users.
The company says the model offers significant improvements in software engineering, scientific research, vision, and long-context reasoning while incorporating safeguards designed for public deployment.
The announcement came alongside the release of Claude Mythos 5, a more permissive version intended for cybersecurity professionals and infrastructure partners through Anthropic’s Project Glasswing initiative.
According to Anthropic, Claude Fable 5 and Claude Mythos 5 are built on the same underlying model. However, Fable 5 includes additional safety mechanisms that redirect certain sensitive requests to Claude Opus 4.8 instead of handling them directly.
The company said these safeguards primarily apply to requests involving cybersecurity, biology and chemistry, and model distillation. Anthropic noted that more than 95% of user sessions do not trigger any fallback mechanisms.
Mythos 5, by contrast, has fewer restrictions and is initially being deployed to select cybersecurity and infrastructure partners through Project Glasswing.
Improved performance across key areas
Anthropic claims Fable 5 outperforms its previous publicly available models on longer and more complex tasks. The company highlighted advancements in software development, knowledge work, scientific research, memory, cybersecurity-related reasoning, and visual understanding.
One of the model’s most notable features is its enhanced vision capability. Anthropic said Fable 5 can extract precise data from detailed scientific charts and figures and can recreate web application source code from screenshots.
The company also revealed that the model successfully completed the video game Pokémon FireRed using only raw screenshots without relying on maps, navigation tools, or external assistance.
Strong results from early testing
Several organizations involved in early testing reported significant productivity gains.
Financial infrastructure company Stripe said Fable 5 completed a codebase-wide migration across a 50-million-line Ruby codebase in a single day. According to the company, the same task would have taken a team of engineers more than two months to complete manually.
Trading firm IMC reported that the model performed well across factual lookup, conceptual reasoning, root-cause analysis, and expected-value analysis evaluations.
Anthropic also cited results from Hebbia’s Finance Benchmark, where Fable 5 achieved the highest score among tested AI models for senior-level financial reasoning tasks.
In frontier physics research, one testing partner reported that Fable 5 reached nearly the same outcome as GPT-5.5 after 36 hours while using roughly one-third of the reasoning tokens. GPT-5.5 reportedly required four days to produce a comparable result.
Memory and long-context enhancements
Anthropic said Fable 5 shows significant improvements in memory and long-context performance.
During testing with the deck-building game Slay the Spire, file-based memory improved Fable 5’s performance three times more than it improved Claude Opus 4.8 under identical conditions.
The company believes these improvements will help users handle larger projects, extended research tasks, and complex workflows more efficiently.
Pricing and availability
Claude Fable 5 is currently available to subscribers on Pro, Max, Team, and seat-based Enterprise plans at no additional cost through June 22.
Beginning June 23, users will need usage credits to access the model unless Anthropic determines that available capacity allows the free-access period to continue.
For developers, Fable 5 is available through Anthropic’s API under the identifier “claude-fable-5.” Pricing has been set at $10 per million input tokens and $50 per million output tokens, making it less than half the cost of Claude Mythos Preview.
Mythos 5 focuses on cybersecurity
Alongside Fable 5, Anthropic introduced Claude Mythos 5, which the company describes as possessing some of the strongest cybersecurity capabilities available in an AI model.
The model is being deployed through Project Glasswing, a collaboration with the U.S. government and selected partners focused on defensive cybersecurity applications. Mythos 5 serves as the successor to Claude Mythos Preview and operates with fewer cybersecurity-related restrictions than its public counterpart.
Data retention policy
Anthropic also announced a new data retention policy for Mythos-class models.
The company said all traffic involving these models will be retained for up to 30 days and will not be used for AI training purposes. Anthropic added that data will be deleted after the retention period in nearly all cases.
.gif)
.jpg)
