Anthropic today announced the release of two new AI models, Claude Fable 5 and Claude Mythos 5, in one of the company’s most significant model releases to date. The launch highlights Anthropic’s commitment to building more capable AI systems. At the same time, the company prioritizes safety, controlled access, and responsible deployment.
Claude Fable 5 is now available for general use, while Claude Mythos 5 is being released to a limited group of trusted users. This group includes cyberdefense organizations and critical infrastructure partners.
What Is Claude Fable 5?
Claude Fable 5 is Anthropic’s newest generally available AI model. According to Anthropic, it is the company’s most capable broadly released model so far. There are major improvements across software engineering, knowledge work, vision, scientific research, and long-context reasoning.
The model is designed to handle longer and more complex tasks than previous Claude models. As a result, Claude Fable 5 is especially relevant for developers, researchers, analysts, and enterprise users. These users rely on AI for multi-step workflows.
Anthropic says Claude Fable 5 performs strongly on benchmark tests and is particularly effective when working on difficult, extended tasks. These tasks require planning, reasoning, and execution over time.
Claude Fable 5 Gains Big in Coding and Software Engineering
One of the biggest gains for Claude Fable 5 is in software engineering. Anthropic says the model can work across large codebases, complete complex migrations, and help with production-level coding tasks.
Early testers have apparently used Claude Fable 5 for sophisticated coding workflows, such as implementing large-scale codebase modifications that generally require substantial engineering effort. In addition, the model performed well on coding evaluations that aim to assess whether AI can handle difficult programming tasks in line with professional software standards.
For developers, this could make Claude Fable 5 useful for: Codebase migrations Debugging Software prototyping Long-horizon coding tasks Agentic development workflows Understanding unfamiliar tools and frameworks The release also points to a broader trend in AI: models are becoming more capable of acting as autonomous engineering assistants rather than simple code-generation tools.
Improvements in Knowledge Work and Analysis
Claude Fable 5 is also positioned as a stronger model for knowledge work. Anthropic says the model is good at analytical tasks with documents, charts, tables, finance and complex reasoning.
For businesses that need AI assistance with research, financial analysis, legal review, operations, and decision support, this might make Claude Fable 5 worth its salt. The model is built to interpret nuanced information. It can reason across documents and produce more reliable outputs for professional workflows.
Anthropic touts improved vision capabilities in Claude Fable 5
Anthropic also emphasized that Claude Fable 5’s vision capabilities have been improved. The model can interpret complex visual inputs, collect information from scientific figures, and even operate based on screenshots.
One example Anthropic shared involved Claude Fable 5 completing Pokémon FireRed using only raw game screenshots, without additional maps or navigation tools. The company presented this as evidence of the model’s improved ability to interpret visual information and act on it over time.
These vision improvements may be useful for tasks such as:
- Reading charts and graphs
- Analyzing screenshots
- Extracting data from images
- Understanding scientific visuals
- Rebuilding interfaces from visual references
- Supporting multimodal workflows
Claude Mythos 5 Is Built for Trusted Access Use Cases
Alongside Claude Fable 5, Anthropic also launched Claude Mythos 5. The company says Mythos 5 is based on the same underlying model as Fable 5 but with certain safeguards lifted for approved users.
Claude Mythos 5 is not broadly available. It is initially being deployed through Project Glasswing in collaboration with the U.S. government. The model is aimed at trusted cyberdefenders and infrastructure providers.
Anthropic says Mythos 5 has especially strong cybersecurity capabilities. Because of the potential risks, access is restricted to selected organizations. These organizations can use the model for defensive and beneficial purposes.
Why Anthropic Added New Safeguards
A major part of the Fable 5 launch is safety. Anthropic says models at this capability level can create serious risks if misused, especially in cybersecurity, biology, chemistry, and model distillation.
To manage these risks, Fable 5 uses new safety classifiers. When a request is flagged as potentially risky, the user may receive a response from Claude Opus 4.8 instead of Claude Fable 5.
Anthropic says this fallback system allows general access to Fable 5 while reducing the risk of misuse. However, the company also acknowledged that the safeguards may catch some harmless requests.
The safeguard areas include:
- Cybersecurity misuse
- Biology and chemistry risks
- Attempts to distill or extract model capabilities
- Jailbreak attempts
- Other high-risk dual-use requests
According to Anthropic, these safeguards trigger in fewer than 5% of sessions on average.
Data Retention Policy for Mythos-Class Models
Anthropic also introduced a new data retention policy for Fable 5, Mythos 5, and future models at similar capability levels. The company says traffic on Mythos-class models will require 30-day retention.
Anthropic says it will not use this data to train new Claude models. Instead, the company will use it for safety-related purposes, such as detecting misuse, improving safeguards, and identifying new jailbreak attempts.
This policy may be important for business users evaluating whether Claude Fable 5 fits their privacy and compliance requirements.
Claude Fable 5 Pricing and Availability
Claude Fable 5 is available through the Claude API and selected Claude plans. Developers can access the model using the Claude API under the model name claude-fable-5.
Anthropic says pricing for both Claude Fable 5 and Claude Mythos 5 is:
- $10 per million input tokens
- $50 per million output tokens
Claude Fable 5 is initially included for Pro, Max, Team, and seat-based Enterprise users at no extra cost through June 22, 2026. Starting June 23, Anthropic says usage will require credits unless capacity allows the company to extend access.
Claude Mythos 5 remains restricted to approved users, including Project Glasswing partners and future trusted access participants.
Why This Launch Matters
The launch of Claude Fable 5 and Claude Mythos 5 shows how quickly frontier AI models are moving toward more autonomous, specialized, and high-impact work.
For everyday users and businesses, Claude Fable 5 may offer stronger performance in coding, research, analysis, and vision-based tasks. For trusted organizations, Claude Mythos 5 could support more advanced cybersecurity and scientific research applications.
At the same time, Anthropic’s rollout shows that safety controls are becoming a central part of frontier AI deployment. Rather than releasing all capabilities without restriction, Anthropic is using a tiered access model. There is broad access for Claude Fable 5, restricted access for Claude Mythos 5, and stronger safeguards for high-risk domains.
Final Thoughts
Claude Fable 5 is one of Anthropic’s most ambitious AI releases so far. It brings major improvements in coding, vision, research, and long-context reasoning while introducing new safeguards for potentially risky tasks.
Claude Mythos 5, meanwhile, represents a more controlled path for deploying powerful AI capabilities in sensitive fields such as cybersecurity and life sciences.
As AI models become more capable, the balance between access, performance, safety, and trust will become increasingly important. Anthropic’s Fable 5 and Mythos 5 launch is a clear example of how leading AI companies are trying to manage that balance.
