DeepSeek 將 V4 Pro 價格永久下調 75%,Xiaomi 的 MiMo V2.5 API 價格最高下砍 99%,加劇中國模型市場的價格戰,並把壓力推向全球。
即使單位 token 價格大跌,較低的推理成本仍可能帶動用量以更快速度增加;但 AI 資本支出與銷售成長之間約 46% 的差距,也凸顯變現與投資回報風險。
What caused the Silicon Data LLM Token Expenditure Index to fall below $1 per million tokens to a record low, how did Chinese AI laboratorieAn editorial illustration of the AI inference market’s price war and expanding demand.
AI 提示詞
Create a landscape editorial hero image for this Studio Global article: What caused the Silicon Data LLM Token Expenditure Index to fall below $1 per million tokens to a record low, how did Chinese AI laboratorie. Article summary: The index fell below $1 because the effective market price of inference was pushed down by cheaper open-weight Chinese models, aggressive API discounting, lower production costs, and competitive cuts by frontier provider. Topic tags: general, general web, user generated, news. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts w
openai.com
Silicon Data 的 LLM Token Expenditure Index 在 2026 年 8 月 31 日降至每百萬 token 97 美分,首次跌破 1 美元,也創下該指數自去年底成立以來的最低紀錄。這個讀數較初夏高點低逾一半。4
但「token 變便宜」不等於「AI 花費正在減少」。這項指數是以用量加權計算市場為每百萬個推理 token 支付的價格,反映的是 AI 智能的單位價格與市場支付意願,而不是所有 AI 支出的總額,也不是 token 使用量本身。278
DeepSeek 將 V4 Pro 價格永久下調 75%,Xiaomi 的 MiMo V2.5 API 價格最高下砍 99%,加劇中國模型市場的價格戰,並把壓力推向全球。
即使單位 token 價格大跌,較低的推理成本仍可能帶動用量以更快速度增加;但 AI 資本支出與銷售成長之間約 46% 的差距,也凸顯變現與投資回報風險。
What caused the Silicon Data LLM Token Expenditure Index to fall below $1 per million tokens to a record low, how did Chinese AI laboratorieAn editorial illustration of the AI inference market’s price war and expanding demand.
AI 提示詞
Create a landscape editorial hero image for this Studio Global article: What caused the Silicon Data LLM Token Expenditure Index to fall below $1 per million tokens to a record low, how did Chinese AI laboratorie. Article summary: The index fell below $1 because the effective market price of inference was pushed down by cheaper open-weight Chinese models, aggressive API discounting, lower production costs, and competitive cuts by frontier provider. Topic tags: general, general web, user generated, news. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts w
openai.com
Silicon Data 的 LLM Token Expenditure Index 在 2026 年 8 月 31 日降至每百萬 token 97 美分,首次跌破 1 美元,也創下該指數自去年底成立以來的最低紀錄。這個讀數較初夏高點低逾一半。4
但「token 變便宜」不等於「AI 花費正在減少」。這項指數是以用量加權計算市場為每百萬個推理 token 支付的價格,反映的是 AI 智能的單位價格與市場支付意願,而不是所有 AI 支出的總額,也不是 token 使用量本身。278