Anthropic has introduced Fable 5.1, an improved version of its Fable 5 model, which shows better performance in various benchmarks. According to the company, Fable 5.1 uses cache reads that cost 75% less than the previous model. This means that for common tasks, users could see a cost reduction of about 25%, and up to 45% for more intensive uses, such as running complex AI agents. Anthropic describes Fable 5.1 and its counterpart, Claude Opus 5, as its most advanced models for tasks like coding and knowledge processing. On specific benchmark tests, Fable 5.1 scores 55.8% on Terminal-Bench 4.0, compared to 42.0% for Fable 5. In a science-focused test called Terminal-Bench-Science 0.1, Fable 5.1 scores 52.6%, which is more than double that of its predecessor. However, the details and methodology of these tests are not publicly available, making it difficult to fully verify the improvements. The key to the cost savings lies in the use of cache—storing and reusing previously processed information to avoid redundant processing and charges. Anthropic has not yet released the full pricing grid for Fable 5.1, which makes it hard to calculate exact costs at the moment. The principle behind the savings is straightforward: using cached information reduces the need to pay for repeated processing of the same data, such as long prompts or repeated code reads. The more a task repeats the same context, the greater the discount. This move by Anthropic aligns with similar strategies from competitors like OpenAI and Google, which also offer reduced pricing for cache reads. In addition to improving performance and cost, Anthropic is targeting enterprise users with the Enterprise Frontier Safeguards (EFS) program. This initiative allows companies to keep their data in their own cloud storage, encrypted with their own keys, without any Anthropic employees accessing it. Another project, Claude Mythos 5.1, is aimed at specialized fields like cybersecurity and life sciences research, with access restricted to approved programs. Anthropic is clearly aiming to reassure both large enterprises and technical teams by focusing on security, pricing, and performance improvements. The full pricing details will be crucial for technical teams to evaluate the model's value, and Anthropic plans to release them in the near future.