AI startup and research company Anthropic has announced the release of Claude Fable 5, a new AI model classified as “Mythos-class” that has been adapted for broad commercial availability. The company described the model as its most capable publicly accessible system to date, delivering significant advances across software engineering, knowledge work, scientific research, vision-based tasks, and long-context reasoning.
According to Anthropic, Claude Fable 5 outperforms previous Claude models on a wide range of industry benchmarks, particularly when handling complex and extended tasks. However, the company noted that the model’s advanced capabilities also introduce potential risks, especially in areas such as cybersecurity. To address these concerns, Anthropic has implemented safety controls that redirect certain sensitive queries to its less capable Claude Opus 4.8 model. The safeguards are designed to limit misuse while allowing broad access, although the company acknowledged that some benign requests may occasionally be affected.
Alongside Fable 5, Anthropic introduced Claude Mythos 5, a version of the same underlying model with selected safeguards removed. Initially, access to Mythos 5 will be limited to cybersecurity organizations and infrastructure providers participating in Project Glasswing, a program developed in collaboration with the U.S. government. Anthropic stated that it plans to expand availability through a broader trusted access framework in the future.
Performance results released by the company highlight major gains in software engineering. During testing, financial technology firm Stripe reported that the model completed a large-scale migration across a 50-million-line codebase in a single day, a task that would typically require months of work by a team of engineers. Anthropic also said Fable 5 achieved leading results in coding evaluations designed to measure performance on complex production-level software tasks.
Knowledge-based reasoning is another area where the model demonstrated strong results. Anthropic reported that Fable 5 achieved top scores on financial and analytical benchmarks, showing improvements in document analysis, chart interpretation, problem-solving, and decision-making tasks. The company also described Fable 5 as a leading vision model capable of extracting detailed information from scientific figures, recreating software applications from screenshots, and handling advanced image-based challenges.
Anthropic highlighted the model’s ability to maintain performance across extremely large contexts and extended workflows. The company said Fable 5 can use persistent memory more effectively than previous Claude models, enabling stronger long-term reasoning and task execution.
The company also emphasized the capabilities of Mythos 5 in scientific research. Internal testing reportedly showed that the model accelerated parts of the drug discovery process by roughly tenfold and independently completed tasks typically performed by human researchers. Anthropic further stated that Mythos 5 generated novel molecular biology hypotheses that researchers frequently preferred over those produced by earlier models. In genomics research, the model reportedly conducted an extended autonomous study involving millions of cells from 138 animal species and developed a machine learning system that outperformed a recently published scientific model despite being significantly smaller.
Anthropic said both Fable 5 and Mythos 5 are priced at $10 per million input tokens and $50 per million output tokens, representing a substantial reduction from the cost of Claude Mythos Preview. Claude Fable 5 is available immediately through Anthropic’s platforms and API, while access to Mythos 5 remains restricted to approved cybersecurity partners and selected research organizations until a broader trusted access program is established.
The post Anthropic Unveils Claude Fable 5 With Enhanced Software Engineering, Vision, And Long-Context Performance appeared first on Metaverse Post.


