en.Wedoany.com Reported - On June 9, U.S. artificial intelligence company Anthropic released its next-generation general-purpose large model, Claude Fable 5, and simultaneously launched Claude Mythos 5 for trusted scenarios. Claude Fable 5 is positioned as a high-capability model open to general users, covering areas such as software engineering, knowledge work, visual understanding, scientific research, and long-task processing.
A key change in this release is that Anthropic is bringing higher-tier model capabilities to public use. Claude Fable 5 is built on Mythos-level capabilities but incorporates stricter safety mechanisms, enabling it to be open to a broader user base. The model shows significant improvements in complex analysis, long-cycle tasks, and code engineering, handling longer and more complex autonomous tasks while demonstrating stronger continuous execution capabilities in document reasoning, chart comprehension, software migration, visual tasks, and multi-step research. For enterprise users and developers, this model is not just a chat tool but more akin to an intelligent work system capable of engineering decomposition, document research, code generation, business analysis, and prototype construction.
The safety mechanism is a key boundary distinguishing Claude Fable 5 from Claude Mythos 5. For requests that may involve cybersecurity attacks, biochemical risks, or model capability distillation, Claude Fable 5 identifies them through classifiers and, in some scenarios, routes them to Claude Opus 4.8 for processing. Claude Mythos 5, on the other hand, is open to more restricted partners and trusted access programs, used for cyber defense, critical software infrastructure protection, and subsequent life science research scenarios. This "dual-model" release approach indicates that frontier AI companies are attempting to split high-capability models into general open versions and controlled professional versions to balance commercial applications, research value, and safety risks.
In terms of commercial use, Claude Fable 5 is already available through the Claude API and select enterprise programs, priced at $10 per million input tokens and $50 per million output tokens. Anthropic has also included it in phased usage windows for Pro, Max, Team, and some enterprise subscription plans, with adjustments based on capacity. For AI application developers, the release of the new model means stronger code generation, agent tasks, and knowledge processing capabilities can be directly integrated into products; for enterprise customers, the model's ability to stably handle complex documents, long-process tasks, and internal tool calls will impact its adoption speed in office, R&D, finance, legal, data analysis, and customer service scenarios.
For the information and communication technology industry chain, the release of Claude Fable 5 will continue to amplify AI infrastructure demand. High-capability models require stronger inference computing power, higher-throughput API platforms, more stable cloud services, finer-grained security auditing, and enterprise-level data governance capabilities. As models can execute longer tasks, the computational load, context length, and tool coordination needs per user call will also increase, driving demand in cloud computing, data centers, AI servers, network transmission, model security, API gateways, and enterprise application integration. Key focus areas ahead include model capacity expansion, changes in enterprise subscription policies, optimization of safety mechanism false positive rates, and the expansion of the Claude Mythos 5 trusted access program. If Claude Fable 5 rapidly gains adoption among developers and enterprises, Anthropic will further strengthen its differentiated position in the general-purpose large model competition, targeting complex tasks and high-security scenarios.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
