Claude Fable 5.1 kullanıma sunuldu: Ajan görevlerinde maliyet yüzde 45’e kadar düşebilecek – Teknoloji Haberleri

Anthropic recently launched Claude Fable 5.1, a model aimed at coding and long‑term information tasks. Compared to its predecessor, the new version promises a roughly 25 % cost reduction on typical workloads and up to 45 % savings when tasks heavily reuse cached data. The savings stem from a 75 % cut in the cache‑read fee, dropping it to $0.25 per million tokens, while the base API rates stay at $10 per million input tokens and $50 per million output tokens.

The model is reported to outperform the earlier version in coding, research, document preparation and sustained agent operations. Test results indicate that Fable 5.1 can identify the root causes of software bugs, not merely their symptoms. Early‑access participants note that the system processes requests faster, consumes fewer tokens, and delivers clearer responses for extended tasks.

Security updates accompany the release. The new safety layer aims to lower the rate of false‑positive blocks for harmless requests and allows the model to scan source code for vulnerabilities in a defensive capacity. High‑risk actions such as penetration testing, exploit development or binary‑file scans are still blocked or routed to other models. Anthropic estimates that the updated system will cut Claude Code session‑level security interventions by about 60 %. Claude Mythos 5.1, which shares the same base architecture but offers looser security limits for cybersecurity and life‑science use, remains restricted to Anthropic’s trusted‑access program. Claude Fable 5.1 is available through Claude’s paid plans, its API, and on Amazon Web Services, Google Cloud and Microsoft Azure.