全球AI Agent與工具

Anthropic 推出 Claude Haiku 5.5：最便宜、最快的小模型，成本降 75%

Introducing Claude Haiku 5.5

來源：Anthropic (官方)・發布 2026-10-07・本站摘要分析（非全文轉載）

📌 一句話啟示：模型成本急降，個人與小企業也能負擔自動化流程，學會多模型分工（大模型決策、小模型執行）是省錢又實用的技能。

摘要分析

Anthropic 於 2026 年 10 月 7 日推出 Claude Haiku 5.5，定位為最便宜、最快且最強的小模型，適合摘要、分類、資料庫查詢、即時客服與瀏覽器操作，並可作為 Opus／Sonnet 5.5 的子代理。10 萬 token 內定價每百萬輸入 0.10、輸出 0.50 美元，平均運行成本較 Haiku 4.5 降約 75%，短請求更省 90%。公布成績含 OSWorld 2.1 達 72.4%（前代 15.7%）、Terminal-Bench 39.2%，並首度提供可調 effort。同時 Sonnet 5.5 快取讀取價格減半，模型已在三大雲平台上線。

Haiku 5.5 定價 0.10／0.50 美元，成本較前代降約 75%
OSWorld 72.4%、Terminal-Bench 39.2%，小模型也能做代理
首款可調 effort 的 Haiku，適合子代理與高量任務
Sonnet 5.5 快取讀取價格減半
三大雲平台同步提供

Anthropic released Claude Haiku 5.5 on October 7, 2026 as its cheapest, fastest and most capable small model, targeting high-volume work such as summarization, classification, database queries, live customer support, browser use, and subagent roles alongside Opus 5.5 and Sonnet 5.5. Pricing is $0.10 per million input tokens and $0.50 output for prompts up to 100K tokens ($0.50/$2.50 above), which Anthropic says cuts average running cost about 75% versus Haiku 4.5 and 90% for the roughly 90% of requests under 100K tokens. Reported benchmarks include 72.4% on OSWorld 2.1 (vs 15.7% for Haiku 4.5), 39.2% on Terminal-Bench 4.0 and 1620 Elo on GDPval-AA. It is the first Haiku with an adjustable effort setting. Anthropic also halved Sonnet 5.5 cache-read pricing to $0.10, added monthly API credits for Max and Team subscribers, and shipped beta SDK support for computer and browser use. The model is available on AWS, Google Cloud and Azure.