Anthropic’s most capable model is no longer behind closed doors. On June 9, 2026, the company released Claude Fable 5 , the first publicly available version of its Mythos-class AI. The model sits above the Opus class in capability and is available broadly through the Claude API, Amazon Bedrock, Google Vertex AI, and Microsoft Foundry. Here’s a quick look at what Claude Fable 5 is, what it can do, where it fits in enterprise work, and where its limits are.
What Is Claude Fable 5? Claude Fable 5 is Anthropic’s first Mythos-class model made safe for general use. It shares the same underlying model as Claude Mythos 5 but adds safety classifiers that restrict responses in areas like offensive cybersecurity, biology, chemistry, and model distillation. When the model declines a request in those areas, it falls back to Claude Opus 4.8 automatically. According to Anthropic, these classifiers trigger in fewer than 5% of sessions on average.
Claude Mythos 5, by contrast, has those classifiers lifted. It remains restricted to approved organizations through Project Glasswing and is not available to the public.
The API model ID for Claude Fable 5 is claude-fable-5.
Key Capabilities Claude Fable 5 is built for long-running, multi-step tasks rather than single-turn exchanges. Its performance gains are most visible on complex, sustained workflows.
1. Software Engineering Stripe reported that Fable 5 compressed months of engineering work into days during early testing. In a 50-million-line Ruby codebase, the model completed a codebase-wide migration in a single day that would have required a full team over two months to do manually.
2. Knowledge Work and Analysis On Hebbia’s Finance Benchmark for senior-level reasoning, Fable 5 scored highest among tested models, with gains in document-based reasoning, chart interpretation, and problem-solving. IMC reported it passed their trading-analysis evaluations across factual lookup, root-cause analysis, and expected-value analysis.
3. Vision Fable 5 is Anthropic’s current top model for vision tasks. It can extract data from detailed scientific figures, reconstruct web app source code from screenshots, and completed the Pokémon FireRed video game using only raw game screenshots with no additional tooling.
4. Long-Context Retention The model supports a 1M-token context window and handles up to 128k output tokens per request. Anthropic’s internal tests showed Fable 5 benefits from persistent file-based memory at a rate three times higher than Opus 4.8 on complex, multi-session tasks.
For teams building on top of these capabilities, Kanerika’s generative AI consulting services help enterprises integrate models like Fable 5 into production-grade architectures.
Technical Specs of Claude Fable 5 Spec Details API Model ID claude-fable-5 Context Window 1M tokens Max Output 128k tokens per request Input Pricing $10 per million tokens Output Pricing $50 per million tokens Prompt Cache Discount 90% off input tokens Availability Claude API, AWS Bedrock, Vertex AI, Microsoft Foundry
Enterprise Use Cases Fable 5 is positioned for work where earlier models stalled. High-value areas include:
Large-scale code refactoring and codebase migrations across millions of lines Deep research synthesis combining multiple sources over extended sessions Document-heavy workflows like contract review, regulatory analysis, and financial modelling Long-horizon agentic tasks where the model plans, executes, and self-corrects autonomously For enterprises running AI agents in production environments, understanding AI agent challenges and governance requirements becomes critical when deploying models of this capability level.
Organizations building RAG architectures on top of frontier models will also find Fable 5’s long-context and vision capabilities relevant for knowledge retrieval at scale.
What Fable 5 Will Not Do The safety classifiers are intentionally conservative. Fable 5 will decline or redirect requests involving:
Offensive cybersecurity techniques Biology or chemistry queries with dual-use potential Model distillation from Anthropic’s models These refusals return stop_reason: “refusal” as an HTTP 200 response, not an error. Developers can configure fallback logic to retry on Opus 4.8 via server-side or client-side fallback options in the API.
Teams that need to understand the difference between what Fable 5 and Mythos 5 can do in high-stakes domains should review Anthropic’s system card directly.
Pricing and Availability Claude Fable 5 launched on June 9, 2026. From launch through June 22, it is included at no additional cost on Pro, Max, Team, and seat-based Enterprise plans. From June 23, accessing it requires usage credits, with Anthropic stating plans to restore it as a standard subscription feature as capacity allows.
On the API, pricing is $10 per million input tokens and $50 per million output tokens, at less than half the price of Claude Mythos Preview. The 90% prompt caching discount applies to input tokens.
For teams assessing whether frontier AI spend aligns with measurable returns, Kanerika’s analysis of generative AI ROI benchmarks covers the metrics that matter.
What Enterprise Teams Should Consider Fable 5 is not a drop-in replacement for earlier Claude models in production integrations. Three integration changes apply:
1. Refusal handling. Build logic for stop_reason: “refusal” responses. These are not errors but intentional safety responses.
2. Token consumption. Fable 5 uses tokens at a significantly higher rate than Opus 4.8. During conversational use, consumption runs roughly double. Workflow mode with high or xhigh effort settings can exhaust daily limits quickly in testing environments.
3. Data retention. Both Fable 5 and Mythos 5 carry 30-day data retention and are not available under zero data retention (ZDR). Teams with strict data residency requirements should factor this in before deployment.
For enterprise AI programs evaluating how models like Fable 5 fit into broader strategy, Kanerika’s AI strategy consulting and agentic AI services offer structured deployment frameworks built around governance and measurable outcomes.
