Anthropic Releases Claude Haiku 5.5: A Small Model With 1M Context Priced at $0.10 per Million Input Tokens
https://www.marktechpost.com/2026/10/07/anthropic-releases-claude-haiku-5-5-a-small-model-with-1m-context-priced-at-0-10-per-million-input-tokens/📌 Haiku 5.5 定價拆解：百萬 token 入門價砍到 1 折
TL;DR：Claude Haiku 5.5 百萬輸入 token 僅 $0.10，但換 tokenizer 後實際省幅要重新算。
官方標價寫著「降價 90%」，但這句話只對一部分請求成立。MarkTechPost 的報導把 Claude Haiku 5.5 的定價規則、部署管道與效能數字拆得更細，值得工程團隊對照自己的流量結構重新算一次成本。
🤔 定位：高流量場景的執行模型
Anthropic 釋出 Claude Haiku 5.5，官方定位是目前最便宜、最快的小模型，主打摘要、上下文壓縮（compaction）、分類與子智慧體任務。模型維持 1M token 的上下文視窗，輸出上限則到 128K token，支援文字與圖片輸入、輸出文字，知識截止日期為 2026 年 6 月。目前已在 Claude API、Amazon Bedrock、Google Cloud、Microsoft Foundry 以及 Claude Platform on AWS 全面上線，批次（batch）作業在 beta 階段支援最多 300K 輸出 token。
🧩 effort 參數下放，但取樣參數被鎖死
Haiku 5.5 是首個支援 effort 調節的 Haiku 模型，自適應思考預設開啟，effort 參數預設為 medium。報導特別提醒兩個容易踩雷的地方：非預設的 temperature、top_p 或 top_k 數值會直接回傳 400 錯誤；新的 tokenizer 對同樣文字的計數，大約比 Haiku 4.5 多出 30%。這兩點都已寫進官方遷移指南。
📊 定價門檻設在 10 萬 token
定價以 10 萬 token 為分界：10 萬 token 以內，輸入每百萬 token $0.10、輸出 $0.50，快取讀取 $0.01、5 分鐘快取寫入 $0.125；超過 10 萬 token，輸入升到 $0.50、輸出升到 $2.50。相較之下，Haiku 4.5 的價格是輸入 $1、輸出 $5。Anthropic 表示大約 90% 的 Haiku 4.5 請求落在 10 萬 token 以內，考慮新 tokenizer 的膨脹效果後，估計 Haiku 5.5 平均成本約便宜 75%，批次處理再額外省 50%。
報導也做了跨模型比價：GPT-6 Luna 在短上下文的定價與 Haiku 5.5 完全相同，但它的高價門檻要到 272K input token 以上才啟動（$0.20 輸入／$0.75 輸出）。換算下來，若提示詞落在 150K token 這種中段區間，GPT-6 Luna 的標價反而更便宜，說明定價門檻的位置差異會直接影響實際帳單，不能只看起步價。
💡 Sonnet 5.5 仍是效能天花板
在報導引用的系統卡數據中，Sonnet 5.5 在所有項目都領先，包括 Terminal-Bench 4.0 的 70.6%。Anthropic 自己也建議複雜 agentic 編碼任務優先選用 Sonnet 5.5 或 Opus 5.5，這再次印證 Haiku 5.5 的定位是高流量、低複雜度的執行型任務，而非全面取代中高階模型。
⚠️ 比價不能只看起步價
這篇報導最實用的地方，是把「10 萬 token 分界」「272K token 分界」這類門檻數字並列呈現，提醒工程團隊在不同廠商之間比價時，必須對照自己實際的提示詞長度分佈，單看起步價格很容易得出錯誤結論。
🎯 實務啟示
升級或選型前，先統計自家請求的 token 長度分佈，看落在哪個定價區間；若提示詞經常超過 10 萬 token，Haiku 5.5 的降價幅度會被稀釋，甚至不如 GPT-6 Luna 的中段定價。同時務必檢查呼叫程式碼是否使用了非預設的 temperature／top_p／top_k，升級前先過一輪遷移指南可以省下不少除錯時間。
🔗 來源
- 標題：Anthropic Releases Claude Haiku 5.5: A Small Model With 1M Context Priced at $0.10 per Million Input Tokens
- 作者／機構：Michal Sutter（MarkTechPost）
- 連結：https://www.marktechpost.com/2026/10/07/anthropic-releases-claude-haiku-5-5-a-small-model-with-1m-context-priced-at-0-10-per-million-input-tokens/
#ClaudeHaiku #Anthropic #LLMPricing #AmazonBedrock #GoogleCloud #AIInfrastructure #ContextWindow #AICostOptimization #APIPricing #CloudAI
原始資料 MarkTechPost · 收集於 2026-10-08
摘要原文
Anthropic has released Claude Haiku 5.5 , its cheapest and fastest small model to date. It targets high-volume work like summaries, compaction, classification and subagent tasks. It keeps a 1M token context window and up to 128K output tokens. Pricing starts at $0.10 per million input tokens and $0.50 per million output tokens. That is 90% below Claude Haiku 4.5 for prompts up to 100K tokens. Is it deployable? Yes, as a hosted API. Haiku 5.5 is generally available on the Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS. Haiku 5.5 is the first Haiku-class model with an adjustable effort setting. Adaptive thinking is on by default, and the effort parameter defaults to medium . It takes text and images, outputs text, and has a June 2026 knowledge cutoff. Batch jobs support up to 300K output tokens in beta. It is important to note two key things. Non-default temperature , top_p or top_k values return a 400 error. The new tokenizer also counts the same text as roughly 30% more tokens than Haiku 4.5. The migration guide covers both. Pricing splits at 100K prompt tokens. Up to 100K, input costs $0.10 and output $0.50 per million. Cache reads cost $0.01 and 5-minute cache writes $0.125. Above 100K, rates rise to $0.50 input and $2.50 output. Haiku 4.5 charged $1 input and $5 output. Anthropic says about 90% of Haiku 4.5 requests stayed under 100K tokens. After adjusting for the tokenizer, it estimates Haiku 5.5 runs about 75% cheaper on average. Batch processing takes another 50% off. GPT-6 Luna lists identical short-context rates. Its higher tier starts only above 272K input tokens, at $0.20 and $0.75. For a 150K-token prompt, Luna is cheaper on list price. All figures below are Anthropic-reported (see the system card ): Sonnet 5.5 still leads every row, including 70.6% on Terminal-Bench 4.0. Anthropic itself recommends Sonnet 5.5 and Opus 5.5 for complex agentic coding. Three workloads fit Haiku 5.5 best: Sources: Anthropic docs , Anthropic announcement , OpenAI model page , OpenAI pricing , Google model page , Google pricing . Standard-tier list prices, verified October 7, 2026. Check out the Technical Details. All credit goes to the researcher of this project. Also, feel free to follow us on Twitter and don’t forget to join our 150k+ML SubReddit and Subscribe to our Newsletter . Wait! are you on telegram? now you can join us on telegram as well. [Sponsored] The web is the one API most agents are missing. Databases, calendars and repos have APIs. The open web mostly doesn’t. The TinyFish MCP server gives any MCP client four tools: TinySearch, TinyFetch (full pages as markdown, JavaScript included), TinyBrowser for logins and forms, and TinyAgent for multi-step jobs. Search and Fetch are free. The post Anthropic Releases Claude Haiku 5.5: A Small Model With 1M Context Priced at $0.10 per Million Input Tokens appeared first on MarkTechPost .
由 claude-code/sonnet 自動生成