Claude Fable 5.1 Details Lower Cache Costs for 1M Tokens
Anthropic has launched Claude Fable 5.1, an AI model designed for demanding reasoning and long-horizon agentic work, including complex coding tasks and extended autonomous research. The model maintains the same base input and output pricing as Claude Fable 5, but significantly reduces the cost of reading prompts. Developers using the API will see prompt cache read costs drop to $0.25 per million tokens. Claude Fable 5.1 features an expanded million-token context window and can generate up to 128,000 output tokens, making it suitable for large codebases and extensive documentation. The model is now available to all Claude API customers and partner platforms under the model ID claude-fable-5-1.
Anthropic's Claude Fable 5.1 release, with its dramatically lower prompt cache read costs of $0.25 per million tokens, presents a clear advantage for Asian developers working on long-running AI applications. This cost reduction, coupled with the model's million-token context window and 128,000 output token capacity, directly benefits companies in markets like Singapore, South Korea, and India that are investing heavily in AI-driven software development and research. The ability to manage complex, multi-step tasks such as large code refactors or in-depth financial document analysis with greater cost efficiency could accelerate AI adoption in enterprise settings across the region. The model's focus on long-horizon agentic work and improved vision capabilities for dense charts and PDFs also suggests a stronger tool for financial institutions and research firms in Asia. However, developers migrating from Claude Fable 5 must note the breaking change regarding forced tool use, which is no longer supported. This requires adjustments to existing workflows, potentially introducing a minor friction point for early adopters in the region.
相关文章
6 则
Frontier AI at a cost: what Anthropic’s Fable 5.1 means for the US-China model race
Our analysis last week explored what Anthropic’s Fable 5.1 means for the US-China model race.
曾依赖三星和SK海力士，中国智能手机制造商转向长鑫存储
商汤科学家称多模态人工智能突破有望在两年内实现
星展银行在新加坡结算系统推出前扩展黄金存储设施
Partior 与 LSEG 瞄准全天候跨境支付结算
