Anthropic 将 Claude Haiku 5.5 token 价格下调 90%，与 GPT-6 Luna 看齐

Anthropic 发布 Claude Haiku 5.5，10 万 token 以下输入价格为 0.10 美元/百万 token，降幅 90%，与 OpenAI 的 GPT-6 Luna 看齐。对企业推理支出和模型路由的影响。

已发布 已更新 3 分钟阅读 ✓ AI 就绪 分享

TL;DR——三句话

  • Anthropic released Claude Haiku 5.5 with a 90% token-price cut for requests below 100,000 tokens, pricing input at $0.10 per million tokens and output at $0.50 per million tokens. [S1]
  • Above 100,000 tokens, Haiku 5.5 input rises to $0.50 and output to $2.50 per million tokens, a 50% reduction from Haiku 4.5 rather than the headline 90%. [S1]
  • The lower-tier rates match OpenAI's GPT-6 Luna across input, output, and cache read/write lines, removing previous price leverage on those workloads. [S1]

摘自 key_takeaways （真实字段，最多 3 条）。事实与分析分别标注。

Anthropic 将 Claude Haiku 5.5 token 价格下调 90%，与 GPT-6 Luna 看齐Editorial cover for a component sourcing brief. Chip motif; headline "Anthropic 将 Claude Haiku 5.5 token 价格下调 90%，与 GPT-6 Luna 看齐". No reported key numbers on this brief, so no number strip is drawn. 1 cited source. Static cover (no animation).Anthropic 将 Claude Haiku 5.5 token 价格下调90%，与 GPT-6 Luna 看齐XAIPART
Cover image generated from data already shown on this page: the category motif, the headline, and reported key numbers. No decorative illustration; no third-party image.

您从 AI 回答跳转而来——这是可验证的原文

AI 摘要可能遗漏了其结论的依据。原始预测、受影响料号及所有引用均在本页，均可追溯至来源。

XAIPART 预测

本篇简报包含可证伪的预测。

跳转至完整预测及其证伪规则 ›
受影响的料号

此报道在目录中无匹配——我们选择不显示，而不是推断匹配。

了解判定方式 ›
需要采购料号？发送请求 → 询价

Reported by VentureBeat on . Summarized and analyzed by XAIPART Editorial.

What happened

Anthropic released Claude Haiku 5.5 with a steep token-price cut aimed at high-volume enterprise workloads such as summarization, classification, database queries, and sub-tasks handled by larger agent systems.S1 The headline figure is a 90% reduction on requests under 100,000 tokens, with input priced at $0.10 per million tokens and output at $0.50 per million tokens, identical to OpenAI's GPT-6 Luna base rates.S1 Anthropic estimates roughly 75% lower total workload cost versus Haiku 4.5 once request-size mix and an updated tokenizer are included.S1

Haiku 5.5 is positioned as a supporting worker for Opus- and Sonnet-class agents, not as the recommended model for demanding coding assignments.S1 Rogo's applied AI lead Alex Wang said the model is accurate enough for revenue-figure lookups and cheap enough to run at volume.S1 Anthropic states that approximately 90% of Haiku 4.5 requests fell under the 100,000-token threshold.S1

Above 100,000 tokens, input rises to $0.50 and output to $2.50 per million tokens, a 50% cut from Haiku 4.5, and caching rates scale accordingly.S1 The launch also reduces Sonnet 5.5 caching charges and adds monthly API credits for Max and Team subscribers.S1 The source notes that pricing for requests of exactly 100,000 tokens and the exact token accounting rule are not disclosed.S1

Key numbers from Anthropic's launch materials
MetricReported valueSource
Haiku 5.5 input price (<100K tokens)$0.10 per million tokens[S1]
Haiku 5.5 output price (<100K tokens)$0.50 per million tokens[S1]
Haiku 5.5 input price (>100K tokens)$0.50 per million tokens[S1]
Haiku 5.5 output price (>100K tokens)$2.50 per million tokens[S1]
Haiku 4.5 baseline input/output$1.00 / $5.00 per million tokens[S1]
Estimated workload cost reduction vs. Haiku 4.5~75%[S1]
Share of Haiku 4.5 requests under 100K tokens~90%[S1]
GPT-6 Luna input/output parity$0.10 / $0.50 per million tokens[S1]
Gemini 3.5 Flash-Lite input/output$0.30 / $2.50 per million tokens[S1]
Gemini 3.8 Flash promotional end dateDec. 31, 2026[S1]

Background timeline

  1. - Anthropic released Claude Haiku 5.5 with revised token pricing and lower Sonnet 5.5 caching charges. [S1]
  2. No prior milestones disclosed by the source.

XAIPART take

Technical impact

The 5x step between Haiku 5.5's lower and upper tiers means workload shaping matters more than headline rates. Applications that batch above the 100,000-token boundary will see savings shrink from 90% to 50%, and the source notes token prices cannot alone determine job cost. XAIPART analysis

Market impact

Parity with GPT-6 Luna compresses procurement leverage on small-context jobs and pushes differentiation toward context thresholds, cache economics, and negotiated enterprise rates. Procurement teams re-evaluating multi-vendor LLM routing now have a hard floor at $0.10/M input. XAIPART analysis

Action

Engineering leads should map request-size distributions before retraining pipelines on Haiku 5.5, since the 100,000-token cliff and the new tokenizer both affect realized spend. Re-quote representative workloads against GPT-6 Luna, Gemini 3.5 Flash-Lite, and Grok 4.3 using the same benchmark prompts. XAIPART analysis

XAIPART Forecast

XAIPART expects at least one competing proprietary vendor to publish a sub-$0.10/M input tier for short-context requests within 9 months of this brief, conditional on Gemini 3.8 Flash promotional rates not dropping below $0.30/M input by ; if no such tier appears before this forecast is wrong. XAIPART analysis

Facts are as reported by the cited source. The design-in, sourcing and forward-looking statements above are XAIPART's analysis, not the source's claims. Unconfirmed: the exact billing rule at the 100,000-token boundary, the token accounting that defines the threshold, and any negotiated enterprise discounts are not disclosed by the source.

Corrections: none at publication. This brief will be updated with a timestamp if the source corrects its report.

Key figures sharing one unit family (values are copyable from the table above).
Key numbers in the same unit familyComparable figures: Haiku 5.5 input price (<100K tokens) $0.10 per; Haiku 5.5 output price (<100K tokens) $0.50 per; Haiku 5.5 input price (>100K tokens) $0.50 per; Haiku 5.5 output price (>100K tokens) $2.50 per; Haiku 4.5 baseline input/output $1.00; GPT-6 Luna input/output parity $0.10.Haiku 5.5 input price (<100K tokens)$0.10 perHaiku 5.5 output price (<100K tokens)$0.50 perHaiku 5.5 input price (>100K tokens)$0.50 perHaiku 5.5 output price (>100K tokens)$2.50 perHaiku 4.5 baseline input/output$1.00GPT-6 Luna input/output parity$0.10

替代料与生命周期

已收录的替代料——仅为目录范围，不代表市场排名
料号关系生命周期依据
— 暂无收录替代料 待补充数据 — 需 match_type 后才可显示任何前代标签

“前代”和“替代”标签仅来自已记录的匹配类型、生命周期状态或来源文本。没有记录匹配时，不显示任何标签。我们自己的商品列表不构成市场参考。

关注要点

可证伪预测——阈值、时间窗口和当前状态。被推翻的预测仍保留可见。
论断阈值发布方状态
XAIPART expects at least one competing proprietary vendor to publish a sub-$0.10/M input tier for short-context requests within 9 months of this brief, conditional on Gemini 3.8 Flash promotional rates not dropping be... no such tier appears before 2027-07-09 Pending

被证明错误的预测仍保留可见并加以标注。我们不会删除它们。

受影响的料号

目录中无匹配。 此事件未映射到 XAIPART 当前上架的任何料号。我们不推断匹配，也不显示占位列表。

各维度成本压力

Source: FRED · Generated · reference output z and composite are algorithm outputs, not a fetch snapshot

Cost pressure by dimension — composite stateDiverging bars for three dimensions around a zero midpoint. Input Cost has no reading; Supply Demand z -1.4711, notable; Fx Energy z 0.2546, neutral. Composite index -9.1, band neutral.0投入成本no reading供需z −1.4711 · notable46 d汇率与能源z +0.2546 · neutral5 dcomposite −9.1 · neutral−100 · −40 · −15 · 0 · +15 · +40 · +100
Underlying values — copyable for verification.
DimensionFRED series IDUnitLagAs of
投入成本PCOPPUSDM, PALUMUSDM, PNICKUSDMUSD/mt52 d
供需PCU334413334413, IPG3344Sindex(1982=100) / index(2017=100)46 d
汇率与能源DEXCHUS, DCOILBRENTEU, DHHNGSPrate / USD/bbl / USD/MMBtu5 d

No free authoritative interface for: gold · silver · rare earth · polysilicon · lithium carbonate · market share — listed rather than approximated.

本文如何生成

过程证据——抓取、来源可靠性检查、模型与校验
本简报的审计记录
Source fetched VentureBeat · venturebeat.com · 4,072 chars
Source sanity (S1) ok — passed before any model call (no s1_ block recorded)
Model / prompt MiniMax-M3 · prompt Blog Agent - News Summary (v2.8-final) · 7,201 tokens · 20.4s · run completed
Gate verdict pass · 6/6 applicable checks passed · spec v2.9 · C10 ratio 0.013134 · C10 rounds 0
Time contract 4 <time datetime> values in this brief — 0 violations (+08:00 or date-only)

上述每一行均由流水线生成，而非人工撰写。

编辑评分

确定性校验结果——纯 PHP 实现，判定过程不涉及模型
校验评分明细 6 项适用检查， 6项通过： 100 %. No hold items or mandatory failures. Gate spec: v2.9 (S0 + S1 + C1–C20) · Source affinity (C20): 0.6697. 适用检查已通过 6 / 6 = 100% · 判定： pass No hold items or mandatory failures Gate spec: v2.9 (S0 + S1 + C1–C20) · Source affinity (C20): 0.6697 是否发布由校验决定。独立审计评分仅供参考。

独立审计（仅供参考，不影响发布）： 51/100 · 等级 C — 详情

相关报道

发现错误？ 告诉我们哪里有问题 ——更正内容会与原文一同发布并记录在案。我们不会悄悄修改文章。

评论

成为第一个评论的人。

来源

  1. Anthropic launches Claude Haiku 5.5 with 90% API price reduction — https://venturebeat.com/technology/anthropic-launches-claude-haiku-5-5-with-90-api-price-reduction-matching-gpt-6-luna · rss · 访问于

事实标注为 S1/S2，并可追溯至所链接的来源。标注为 分析 的内容是 XAIPART 自身的推理，与报道的事实分开呈现。来源链接会被访问核实，而非 nofollow ——署名是有意为之。