Claude Haiku 5.5
https://simonwillison.net/2026/Oct/7/claude-haiku-5-5/📌 【Anthropic】Haiku 5.5 打平對手價格，tokenizer 卻偷偷漲稅
TL;DR：Haiku 5.5 表面價格追平 GPT-6 Luna，但新 tokenizer 讓同一份 prompt 多吃 1.25 倍 token。
🎣 一邊喊著降價，一邊換上更「吃字」的 tokenizer，帳面上的優惠未必是實際到手的優惠。Simon Willison 實測後發現，這次 Anthropic 的定價策略比表面數字複雜得多。
🤔 跟 GPT-6 Luna 正面對決的定價戰
Haiku 4.5 上市將近一年，定價 $1/百萬 input token、$5/百萬 output token，即便在發布當時也不便宜，更是上個月才推出的 OpenAI GPT-6 Luna 的 10 倍價格。新版 Haiku 5.5 直接把價格砍到跟 Luna 一樣：$0.10/$0.50（百萬 token），上限為 100,000 token。超過這個門檻，價格會跳升 5 倍到 $0.50/$2.50。相較之下，Luna 要到 272,000 token 才會漲價，而且只漲到 $0.20/$0.75。
| 模型 | Input（≤100K） | Output（≤100K） | 超過閾值後 |
|---|---|---|---|
| Claude Haiku 5.5 | $0.10/M | $0.50/M | $0.50/$2.50（>100K） |
| GPT-6 Luna | $0.10/M | $0.50/M | $0.20/$0.75（>272K） |
| Claude Haiku 4.5（舊版） | $1/M | $5/M | 不適用 |
結論很直接：workload 若能塞進 100,000 token 以內，Haiku 5.5 價格打平 Luna，且 Anthropic 宣稱 benchmark 分數更高；一旦超過這個門檻，Luna 明顯更划算。
💡 隱藏的漲價：新 tokenizer 更「吃字」
Simon 用自己開發的 Claude Token Counter 工具測試，同一段長 prompt 在 Haiku 5.5 下，耗用的 token 數量大約是 Haiku 4.5 的 1.25 倍。換句話說，即使單位價格打平對手，實際帳單可能因為 tokenizer 效率變差而悄悄墊高，這是只看牌價容易忽略的細節。
📊 Pelican 測試：reasoning 等級的時間與成本
Haiku 5.5 不支援關閉 reasoning，預設為 medium。Simon 用經典的「畫一隻騎腳踏車的鵜鶘」SVG 測試跑過 low、medium、high、xhigh、max 五個等級，除了 low 以外都畫出了不錯的車架。
| Reasoning 等級 | 花費 | 耗時 |
|---|---|---|
| Low | 0.0936 美分 | 7 秒 |
| Max | 3.3826 美分 | 5 分 9 秒 |
| Haiku 4.5（無 reasoning 選項） | 0.7583 美分 | — |
值得一提的是，max 等級的 reasoning trace 一開頭就寫著「This is the classic pelican-on-bicycle SVG test…」，顯示模型已經「認出」這是業界常見的 benchmark 題目。對照一年前的 Haiku 4.5，當時完全不支援 reasoning 等級，畫出來的鵜鶘品質也明顯較差。
💡 Sonnet 5.5 快取降價，訂閱戶多了 API 額度
同一波更新裡，Anthropic 也把 Sonnet 5.5 的 cache read 價格砍半，並宣布從本週起為 Max 與 Team 訂閱戶加發每月 API 額度：Max 5x 用戶 $100、Max 20x 用戶 $200，Team 訂閱戶最高 $500（在團隊內共用）。額度剛好等於訂閱費本身，且不會累積到下個月，用不完就作廢。使用者可以在 Settings → Billing 選擇要套用額度的 API 組織，也能關閉 auto-reload，讓額度用完後 API 直接停止呼叫，避免帳單爆炸。
⚠️ 超過 100K token 要三思
這次更新對小型、短 prompt 的 workload 相對友善，但長 context 的使用情境（例如大量文件分析）反而要留意：Haiku 5.5 超過 100K token 後漲幅是 Luna 的 2.5 倍，加上 tokenizer 本身更吃字，實際成本差距可能比牌價顯示得更大。
🎯 實務啟示
選型時別只看牌價，先用實際 prompt 跑一次 token 計算器，確認 tokenizer 差異不會吃掉表面上的降價優惠；同時評估 workload 是否會頻繁超過 100K token 門檻，再決定 Haiku 5.5 或 GPT-6 Luna 更划算。若已是 Max 或 Team 訂閱戶，記得去 Billing 頁面把每月 API 額度領出來用。
🔗 來源
- 標題：Claude Haiku 5.5
- 作者／機構：Simon Willison
- 連結：https://simonwillison.net/2026/Oct/7/claude-haiku-5-5/
#Anthropic #ClaudeHaiku #LLMPricing #GPT6Luna #TokenCounter #AIModels #PromptEngineering #LLMBenchmark #APIpricing #AIAgent
原始資料 Simon Willison · 收集於 2026-10-15
摘要原文
As previously promised , here’s Anthropic’s new fast, low cost model: Introducing Claude Haiku 5.5 . The previous Haiku, 4.5, was very much showing its age. It came out almost a year ago , and was priced at $1/million input and $5/million output—relatively expensive even back then, and a full 10x the price of OpenAI’s GPT-6 Luna , released last month. The new Haiku exactly matches the price of GPT-6 Luna—$0.10/$0.50—up to 100,000 tokens. Beyond 100,000 tokens the price increases 5x to $0.50/$2.50. Luna itself has a price increase at 272,000 tokens but only to $0.20/$0.75. Haiku 5.5 also uses a new, less generous tokenizer. My Claude Token Counter tool shows that the same long prompt uses around 1.25x as many tokens with Haiku 5.5 compared to Haiku 4.5, so there’s a hidden price increase there. If your workloads fit in 100,000 tokens, Haiku is the same price as Luna and reports higher benchmark scores. Above 100,000 tokens, Luna looks like a much better deal. The most recent release of llm-anthropic finally fixed it so I don’t need to ship a new version of that plugin for every new model. I tested the new model like this: Here are pelicans for low, medium, high, xhigh, and max . The new Haiku doesn’t let you disable reasoning, and defaults to medium . I got a good bicycle frame for everything beyond low . The low effort pelican cost 0.0936 cents and took 7 seconds. This max effort pelican (with a reasoning trace that starts “This is the classic pelican-on-bicycle SVG test...”) took 5 minutes 9 seconds to generate, but still only cost me 3.3826 cents : (Since the reasoning trace exhibits awareness of the benchmark, here’s Generate an SVG of an armadillo in fishnet tights jaywalking on Mars (on xhigh) , and the same prompt against some other recent models . Background on that .) For comparison, here’s the pelican I got a year ago from Haiku 4.5 (for 0.7583 cents —Haiku 4.5 did not support reasoning levels). It sucked at drawing pelicans: In addition to Haiku 5.5, Anthropic announced today that they are halving the price of cache reads for Sonnet 5.5. They’ve also added API credits to subscription plans: Second, this week, we’ll roll out a new monthly API credit to all Max and Team subscribers for use on the Claude Platform . Max 5x users will get $100 in credits per month, Max 20x users will get $200, and Team subscribers will receive up to $500, pooled across their users. Claiming this is pleasantly easy: navigate to Settings -> Billing and select the API organization that should benefit from the credits every month: The API credits exactly match the cost of the subscription itself. This is really generous—it makes it much easier for subscribers to use the API. Anthropic also let you disable auto-reload for the API, with the consequence that “API requests will stop when your balance runs out”—exactly what you want if you’re planning to burn through those API credits without risk of a nasty billing surprise . Note that the monthly credits do not roll over —use them or lose them. OpenAI still allow you to use your Codex subscription for personal API use, which works out as a better deal for heavy API users. This new credit scheme goes at least some way to overcoming that difference. This is Claude Haiku 5.5 by Simon Willison, posted on 7th October 2026 . Next: A new feature for my blog, built using my voice Previous: We're going to need default hard budget caps on pretty much everything Sponsor me for $10/month and get a curated email digest of the month's most important LLM developments.
由 claude-code/sonnet 自動生成