- Data snapshot: 2026-10-08 · 6 providers · AA v4.3.2 · No tracking
+ Data snapshot: 2026-10-10 · 6 providers · AA v4.3.2 · No tracking
---
@@ -137,6 +137,11 @@ Artificial Analysis 用于跨厂商能力与单任务成本的统一比较。
## 2026-10 快照变化
+- **Google AI Ultra（已于 2026-05 宣布，本次修正旧快照）**：不再是单一 $249.99/月；美国官网现列 **Ultra 5x $99.99/月**、**Ultra 20x $199.99/月**，分别提供约 AI Pro 的 5x / 20x Gemini App 与 Antigravity 用量。Google AI Plus 为 $4.99/月、Google AI Pro 为 $19.99/月（美国官网月价）。
+- **Gemini App 个人账号（2026-10-09 起逐步实施）**：Google 专项通知列出 Free 手动可选 Flash-Lite；AI Plus 可选 Flash-Lite / Flash、不再提供手动 Pro；AI Pro 与 Ultra 可选 Flash-Lite / Flash / Pro。AI Plus 生效时间由 Google 单独邮件通知。*通用帮助页仍描述 Auto 自动模型路由/动态访问，故“手动不可选”不等于证明后台绝不使用高级模型；此限制也不等同 Gemini Developer API 政策。*
+- **Gemini reasoning 档位**：官方 Gemini App 帮助文档列出 Low / Medium / High，可手动选择时由账户权限与模型决定。高 thinking level 可能占用更多订阅额度；这是 **Gemini App 产品设置**，未当作独立的 API 模型 benchmark 点。
+- **来源**：[Google AI 订阅官网](https://gemini.google.com/advanced) · [Google 2026 I/O 官方公告](https://blog.google/products-and-platforms/products/google-one/google-ai-subscriptions/) · [Google Gemini 访问政策](https://support.google.com/gemini/answer/17004136?hl=en) · [通用额度文档](https://support.google.com/gemini/answer/16275805?hl=en)
+
- **Claude Haiku 5.5（10/7）**：新增 Low / Medium / High / XHigh / Max 五个 AA 实测点；AA Index 29 → 43，AA task cost $0.02 → $0.21。官方定价（≤100K prompt）为 $0.10/M input、$0.50/M output、$0.01/M cache read。
- **Claude Sonnet 5.5（10/7）**：Anthropic 将 cache read 从 $0.20/M 下调到 $0.10/M；按当前 AA 统一口径刷新后，Low / Medium / High / XHigh / Max task cost 约为 $0.35 / $0.48 / $0.88 / $2.01 / $5.46。
- **Claude Max / Team（10/7）**：Max 5x / 20x 新增 $100 / $200 每月 Claude Platform API credit；Team 按席位发放并共享池化，最高 $500/月。该 credit 不能用于 Claude 或 Claude Code 的额外交互用量。
@@ -146,7 +151,7 @@ Artificial Analysis 用于跨厂商能力与单任务成本的统一比较。
- **Gemini 4 Argon**：继续以 limited rollout / preview 状态纳入，不视为全面可用。
- **GPT-6 Sol**：保留历史对比，但默认隐藏，由 GPT-6.1 Sol 取代当前位置。
-**推荐结论未变化**：前沿性价比仍为 **GPT-6.1 Sol · Medium**；绝对能力仍为 **Claude Opus 5.5 · Max**。Plan 分类不改，但 Claude Max 5x / 20x 因新增 API credit 的实际价值提高。
+**推荐结论未变化**：前沿性价比仍为 **GPT-6.1 Sol · Medium**；绝对能力仍为 **Claude Opus 5.5 · Max**。原有跨厂商 Plan 推荐不变；Google 的新 Ultra 5x 现在提供介于 AI Pro 与最高 Ultra 之间的重度使用选项。Claude Max 新增 API credit 价值仍保留。
更完整的来源链接在网页的 **Methodology & Sources** 区域以及 `data.js` 中。
diff --git a/data.js b/data.js
index 9e49340..1c9d7c5 100644
--- a/data.js
+++ b/data.js
@@ -1,4 +1,4 @@
-window.MODEL_DATA={updated:"2026-10-08",baseline:{model:"GPT-6 Luna",effort:"max",costPerTask:.07},correction:{none:1,low:1,medium:1,high:1.05,xhigh:1.10,max:1.18,default:1},providerColors:{OpenAI:"#4f7cff",Anthropic:"#e28a52",Google:"#44a774",DeepSeek:"#8b6de9","Z.ai":"#e2b24f",Kimi:"#d65f91"},models:[
+window.MODEL_DATA={updated:"2026-10-10",baseline:{model:"GPT-6 Luna",effort:"max",costPerTask:.07},correction:{none:1,low:1,medium:1,high:1.05,xhigh:1.10,max:1.18,default:1},providerColors:{OpenAI:"#4f7cff",Anthropic:"#e28a52",Google:"#44a774",DeepSeek:"#8b6de9","Z.ai":"#e2b24f",Kimi:"#d65f91"},models:[
{provider:"OpenAI",family:"GPT-6 Luna",effort:"none",intelligence:18,cost:.01,measured:true,status:"current"},{provider:"OpenAI",family:"GPT-6 Luna",effort:"low",intelligence:22,cost:.0045,measured:true,status:"current"},{provider:"OpenAI",family:"GPT-6 Luna",effort:"medium",intelligence:30,cost:.02,measured:true,status:"current"},{provider:"OpenAI",family:"GPT-6 Luna",effort:"high",intelligence:33,cost:.03,measured:true,status:"current"},{provider:"OpenAI",family:"GPT-6 Luna",effort:"xhigh",intelligence:35,cost:.04,measured:true,status:"current"},{provider:"OpenAI",family:"GPT-6 Luna",effort:"max",intelligence:38,cost:.07,measured:true,status:"current",featured:"efficiency"},
{provider:"OpenAI",family:"GPT-6 Sol",effort:"none",intelligence:29,cost:.33,measured:true,status:"deprecated"},{provider:"OpenAI",family:"GPT-6 Sol",effort:"low",intelligence:34,cost:.13,measured:true,status:"deprecated"},{provider:"OpenAI",family:"GPT-6 Sol",effort:"medium",intelligence:40,cost:.25,measured:true,status:"deprecated"},{provider:"OpenAI",family:"GPT-6 Sol",effort:"high",intelligence:42,cost:.37,measured:true,status:"deprecated"},{provider:"OpenAI",family:"GPT-6 Sol",effort:"xhigh",intelligence:44,cost:.52,measured:true,status:"deprecated"},{provider:"OpenAI",family:"GPT-6 Sol",effort:"max",intelligence:48,cost:1.04,measured:true,status:"deprecated"},
{provider:"OpenAI",family:"GPT-6.1 Sol",effort:"low",intelligence:42,cost:.13,measured:true,status:"current"},{provider:"OpenAI",family:"GPT-6.1 Sol",effort:"medium",intelligence:48,cost:.21,measured:true,status:"current",featured:"frontier-value"},{provider:"OpenAI",family:"GPT-6.1 Sol",effort:"high",intelligence:50,cost:.32,measured:true,status:"current"},{provider:"OpenAI",family:"GPT-6.1 Sol",effort:"xhigh",intelligence:51,cost:.39,measured:true,status:"current"},{provider:"OpenAI",family:"GPT-6.1 Sol",effort:"max",intelligence:52,cost:.72,measured:true,status:"current"},
@@ -15,8 +15,8 @@ window.MODEL_DATA={updated:"2026-10-08",baseline:{model:"GPT-6 Luna",effort:"max
plans:[
{provider:"OpenAI",plan:"ChatGPT Plus",price:"$20/mo",bestFor:"日常全能 / Chat / Work / Research",note:"Oct 7 起 Chat 使用 GPT-6 Sol；Work/Codex 含有限 Astra 用量，具体限额动态变化。",official:true},{provider:"OpenAI",plan:"ChatGPT Pro 100",price:"$100/mo",bestFor:"重度 Chat / Work / Codex",note:"Chat 可用 GPT-6 Pro（Astra）；Work/Codex 高于 Plus 的持续额度。",official:true,recommended:true},{provider:"OpenAI",plan:"ChatGPT Pro 200",price:"$200/mo",bestFor:"更高持续用量",note:"新订阅已重新开放；官方注明非 grandfathered 新订阅的内含用量低于旧档。",official:true},{provider:"OpenAI",plan:"ChatGPT Pro 500",price:"$500/mo",bestFor:"极重度 / Astra Ultrafast",note:"Pro 档最高内含用量；唯一自助 Pro 档包含 Astra Ultrafast。",official:true},
{provider:"Anthropic",plan:"Claude Pro",price:"$20/mo",bestFor:"Claude / Claude Code 轻中度",note:"Oct 7 新增的通用 API 月度额度不覆盖 Pro；API 仍另计。",official:true},{provider:"Anthropic",plan:"Claude Max 5x",price:"$100/mo",bestFor:"高质量编码 / Opus",note:"约 5× Pro session allowance；新增 $100/月 Claude Platform API credit（不用于 Claude/Claude Code 额外用量）。",official:true,recommended:true},{provider:"Anthropic",plan:"Claude Max 20x",price:"$200/mo",bestFor:"Claude 重度日用",note:"约 20× Pro session allowance；新增 $200/月 Claude Platform API credit。",official:true},
-{provider:"Google",plan:"Google AI Pro",price:"$19.99/mo (US)",bestFor:"Gemini + Google 生态",note:"地区价格和权益不同。",official:true},{provider:"Google",plan:"Google AI Ultra",price:"$249.99/mo (US)",bestFor:"Google 最高限额 / Deep Think",note:"最高使用限额；Gemini 4 Argon 将优先从付费 API 与 Ultra 用户开始扩展。",official:true},
+{provider:"Google",plan:"Gemini Free (无 AI Plan)",price:"$0/mo",bestFor:"基础 Gemini 使用",note:"2026-10-09 起个人账户开始调整：专项通知中的手动模型权限为 Flash-Lite；通用说明仍提及自动路由，不能等同后台绝不调用 Flash/Pro。非 Gemini Developer API 政策。",official:true},{provider:"Google",plan:"Google AI Plus",price:"$4.99/mo (US)",bestFor:"低预算 Gemini / Flash",note:"Gemini Apps 个人账户：Flash-Lite + Flash；专项通知取消手动选择 Pro，实际生效日以 Google 发给订户的邮件为准。Gemini 用量约为免费档 2x。",official:true},{provider:"Google",plan:"Google AI Pro",price:"$19.99/mo (US)",bestFor:"Gemini Pro + Google 生态 / Coding",note:"Gemini Apps 可选 Flash-Lite、Flash、Pro；用量约为免费档 4x，含更多 Jules / Gemini CLI / Antigravity 权益。价格与地区资格以结算页为准。",official:true},{provider:"Google",plan:"Google AI Ultra 5x",price:"$99.99/mo (US)",bestFor:"重度 Gemini / Antigravity / Agent",note:"2026-05 Google I/O 新增 Ultra 5x：Gemini Apps 与 Antigravity 约为 AI Pro 的 5x 使用额度；美国官网定价。",official:true},{provider:"Google",plan:"Google AI Ultra 20x",price:"$199.99/mo (US)",bestFor:"最高额 Gemini / Deep Think / Agent",note:"原约 $249.99/月 Ultra 主档已降至 $199.99；Gemini Apps 与 Antigravity 约为 AI Pro 的 20x 使用额度。Gemini 4 Argon 仍按 limited rollout 标注。",official:true},
{provider:"Z.ai",plan:"GLM Coding Lite",price:"$12.6/mo promo",bestFor:"最低价专用 Coding",note:"促销价可能变化。",official:true,recommended:true},{provider:"Z.ai",plan:"GLM Coding Pro",price:"$56/mo promo",bestFor:"日常中型 repo",note:"促销价可能变化。",official:true},{provider:"Z.ai",plan:"GLM Coding Max",price:"$117.6/mo promo",bestFor:"GLM 重度开发",note:"峰值专属资源。",official:true},
{provider:"Kimi",plan:"Andante",price:"¥49/mo",bestFor:"轻量 Agent / Kimi Code",note:"轻量档。",official:true},{provider:"Kimi",plan:"Moderato",price:"¥99/mo",bestFor:"国内综合性价比 / 频繁编码",note:"Code + Agent 综合档。",official:true,recommended:true},{provider:"Kimi",plan:"Allegretto",price:"¥199/mo",bestFor:"专业 Agent",note:"更多 Agent 配额。",official:true},{provider:"Kimi",plan:"Allegro",price:"¥699/mo",bestFor:"最高额度 / 长上下文",note:"最高项目配额。",official:true}],
sources:[
-{label:"Artificial Analysis — GPT-6.1 Sol",url:"https://artificialanalysis.ai/models/releases/gpt-6-1-sol"},{label:"Artificial Analysis — GPT-6 Luna",url:"https://artificialanalysis.ai/models/releases/gpt-6-luna"},{label:"Artificial Analysis — GPT-6 Astra",url:"https://artificialanalysis.ai/models/releases/gpt-6-astra"},{label:"Artificial Analysis — Claude Opus 5.5",url:"https://artificialanalysis.ai/models/releases/claude-opus-5-5"},{label:"Artificial Analysis — Claude Sonnet 5.5",url:"https://artificialanalysis.ai/models/releases/claude-sonnet-5-5"},{label:"Anthropic — Claude Haiku 5.5",url:"https://www.anthropic.com/claude-haiku-5-5"},{label:"Artificial Analysis — Claude Haiku 5.5",url:"https://artificialanalysis.ai/models/releases/claude-haiku-5-5"},{label:"Anthropic — Claude Fable 5.1 / Mythos 5.1",url:"https://www.anthropic.com/claude-fable-and-mythos-5-1"},{label:"Artificial Analysis — Gemini 4 Argon",url:"https://artificialanalysis.ai/models/gemini-4-argon"},{label:"Google — Gemini 4 Argon",url:"https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/"},{label:"Artificial Analysis — Gemini 3.8 Flash",url:"https://artificialanalysis.ai/models/releases/gemini-3-8-flash"},{label:"Artificial Analysis — DeepSeek V4.1 Flash",url:"https://artificialanalysis.ai/models/releases/deepseek-v4-1-flash"},{label:"Artificial Analysis — GLM-5.3 Flash",url:"https://artificialanalysis.ai/models/glm-5-3-flash"},{label:"Artificial Analysis — Kimi K3",url:"https://artificialanalysis.ai/models/comparisons/kimi-k3-low-vs-kimi-k3"},{label:"OpenAI — ChatGPT release notes",url:"https://help.openai.com/en/articles/6825453-chatgpt-release-notes"},{label:"OpenAI — ChatGPT Plus",url:"https://help.openai.com/en/articles/6950777-what-is-chatgpt-plus"},{label:"OpenAI — ChatGPT Pro tiers",url:"https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"},{label:"Anthropic — Max plan",url:"https://support.claude.com/en/articles/11049741-what-is-the-max-plan"},{label:"Anthropic — Max / Team monthly API credits",url:"https://support.claude.com/en/articles/17154008-monthly-api-credits-for-max-and-team-plans"},{label:"Google — AI plans",url:"https://one.google.com/about/google-ai-plans/"},{label:"DeepSeek — Thinking Mode",url:"https://api-docs.deepseek.com/guides/thinking_mode/"},{label:"Z.ai — GLM Coding",url:"https://zcode.z.ai/en"},{label:"Kimi — Membership",url:"https://www.kimi.com/en/help/membership/membership-overview"}]};
\ No newline at end of file
+{label:"Artificial Analysis — GPT-6.1 Sol",url:"https://artificialanalysis.ai/models/releases/gpt-6-1-sol"},{label:"Artificial Analysis — GPT-6 Luna",url:"https://artificialanalysis.ai/models/releases/gpt-6-luna"},{label:"Artificial Analysis — GPT-6 Astra",url:"https://artificialanalysis.ai/models/releases/gpt-6-astra"},{label:"Artificial Analysis — Claude Opus 5.5",url:"https://artificialanalysis.ai/models/releases/claude-opus-5-5"},{label:"Artificial Analysis — Claude Sonnet 5.5",url:"https://artificialanalysis.ai/models/releases/claude-sonnet-5-5"},{label:"Anthropic — Claude Haiku 5.5",url:"https://www.anthropic.com/claude-haiku-5-5"},{label:"Artificial Analysis — Claude Haiku 5.5",url:"https://artificialanalysis.ai/models/releases/claude-haiku-5-5"},{label:"Anthropic — Claude Fable 5.1 / Mythos 5.1",url:"https://www.anthropic.com/claude-fable-and-mythos-5-1"},{label:"Artificial Analysis — Gemini 4 Argon",url:"https://artificialanalysis.ai/models/gemini-4-argon"},{label:"Google — Gemini 4 Argon",url:"https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/"},{label:"Artificial Analysis — Gemini 3.8 Flash",url:"https://artificialanalysis.ai/models/releases/gemini-3-8-flash"},{label:"Artificial Analysis — DeepSeek V4.1 Flash",url:"https://artificialanalysis.ai/models/releases/deepseek-v4-1-flash"},{label:"Artificial Analysis — GLM-5.3 Flash",url:"https://artificialanalysis.ai/models/glm-5-3-flash"},{label:"Artificial Analysis — Kimi K3",url:"https://artificialanalysis.ai/models/comparisons/kimi-k3-low-vs-kimi-k3"},{label:"OpenAI — ChatGPT release notes",url:"https://help.openai.com/en/articles/6825453-chatgpt-release-notes"},{label:"OpenAI — ChatGPT Plus",url:"https://help.openai.com/en/articles/6950777-what-is-chatgpt-plus"},{label:"OpenAI — ChatGPT Pro tiers",url:"https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"},{label:"Anthropic — Max plan",url:"https://support.claude.com/en/articles/11049741-what-is-the-max-plan"},{label:"Anthropic — Max / Team monthly API credits",url:"https://support.claude.com/en/articles/17154008-monthly-api-credits-for-max-and-team-plans"},{label:"Google — Gemini plan pricing (US)",url:"https://gemini.google.com/advanced"},{label:"Google — AI Ultra 5x / 20x official announcement",url:"https://blog.google/products-and-platforms/products/google-one/google-ai-subscriptions/"},{label:"Google — Gemini model access changes (Oct 2026)",url:"https://support.google.com/gemini/answer/17004136?hl=en"},{label:"Google — Gemini model selection / reasoning levels",url:"https://support.google.com/gemini/answer/18648722?hl=en"},{label:"Google — Gemini usage limits",url:"https://support.google.com/gemini/answer/16275805?hl=en"},{label:"Google — AI plans",url:"https://one.google.com/about/google-ai-plans/"},{label:"DeepSeek — Thinking Mode",url:"https://api-docs.deepseek.com/guides/thinking_mode/"},{label:"Z.ai — GLM Coding",url:"https://zcode.z.ai/en"},{label:"Kimi — Membership",url:"https://www.kimi.com/en/help/membership/membership-overview"}]};
\ No newline at end of file
diff --git a/index.html b/index.html
index 5e8dc37..1c8d1bc 100644
--- a/index.html
+++ b/index.html
@@ -17,7 +17,7 @@
-
FRONTIER MODEL EFFICIENCY MAP · 2026-10-08
+
FRONTIER MODEL EFFICIENCY MAP · 2026-10-10
模型 × Reasoning：能力 / 消耗 / Plan
@@ -138,10 +138,11 @@
只补齐官方存在但 AA 未测的档位
纯编码最低门槛GLM Coding Lite$12.6/月促销价，10k credits/week
国内综合Kimi Moderato¥99/月，Code + Agent + 2 并行
+
Google 订阅已按官方重整：AI Plus $4.99、AI Pro $19.99、AI Ultra 5x $99.99、Ultra 20x $199.99（美国月费）。2026-10-09 起个人 Gemini App 模型权限陆续调整；手动模型选择、自动路由与 Developer API 额度须分别看待。
