Anthropic 发布具备代理能力的中端模型 Claude Sonnet 5,Fable 5 获批但未同步推出
英文摘要
Anthropic launched Claude Sonnet 5 as its new default mid-tier model with a 1M-token context window, pricing at $3/$15 per million input/output tokens (promotional $2/$10 until Aug/Sept). Independent benchmarks show meaningful gains over Sonnet 4.6 on coding and agentic tasks (e.g., CursorBench 57% vs 49%, FrontierCode Extended 53.8% score) but still below Opus 4.8 on broad intelligence. However, a tokenizer change and higher turn-taking in evaluations make effective per-task costs sometimes higher than Opus 4.8. Fable 5 was approved for re-release after government engagement but was not launched, leading to a wave of disappointment and speculation. The coding-agent ecosystem (Cursor, Devin, Cline, etc.) adopted Sonnet 5 rapidly, treating it as a practical workhorse model.
中文摘要
Anthropic 发布 Claude Sonnet 5 作为默认中端模型,具备 100 万 token 上下文窗口,定价为输入/输出每百万 token 3/15 美元(8-9 月促销价 2/10 美元)。第三方基准测试显示其在编程和代理任务上较 Sonnet 4.6 有显著提升(如 CursorBench 57% vs 49%,FrontierCode Extended 得分 53.8%),但在通用智能上仍低于 Opus 4.8。然而,分词器变化和评估中更多轮次导致实际每任务成本有时高于 Opus 4.8。Fable 5 在政府沟通后获批重新发布但并未推出,引发失望和猜测。编程代理生态(Cursor、Devin、Cline 等)迅速采用 Sonnet 5,将其视为生产环境中的实用主力模型。
关键要点
Claude Sonnet 5 is officially released with 1M-token context, standard $3/$15 pricing, and a promotional rate of $2/$10 through Aug/Sept.
Claude Sonnet 5 正式发布,提供 100 万 token 上下文,标准定价为 3/15 美元,8-9 月促销价 2/10 美元。
Sonnet 5 shows solid coding and agentic improvements over Sonnet 4.6 but does not surpass Opus 4.8 on general intelligence benchmarks.
Sonnet 5 在编程和代理能力上较 Sonnet 4.6 有明显提升,但在通用智能基准上未超越 Opus 4.8。
Due to tokenizer changes and more turn-taking, effective per-task cost can exceed Opus 4.8, dampening excitement despite list-price parity or improvement.
由于分词器变化和更多轮次,实际每任务成本可能高于 Opus 4.8,这削弱了低标价带来的兴奋感。
Fable 5 was not released concurrent with Sonnet 5, leading to user disappointment and speculation about regulatory or strategic gating.
Fable 5 未与 Sonnet 5 同步发布,导致用户失望,并引发关于监管或策略性限发的猜测。
Leading coding tools (Cursor, Devin, Cline) integrated Sonnet 5 immediately, signaling its role as a practical production-grade agent model.
主流编程工具(Cursor、Devin、Cline)立即集成 Sonnet 5,表明其作为生产级代理模型的实用定位。