All Products
Search

This Product

    E-MapReduce:AI function pricing

    Document Center

    E-MapReduce:AI function pricing

    Last Updated:Sep 10, 2026

    AI Function is a built-in SQL AI inference capability in EMR Serverless StarRocks that lets you call qwen-series LLMs directly from SQL. This topic describes its billing rules, per-region per-model token pricing, and fee calculation examples. For available AI functions, invocation syntax, and related configuration items, see AI center.

    Billing

    Note

    Each user receives a free quota of 1 million tokens for the first month. Any usage that exceeds this free quota is billed at standard rates.

    Feature

    Description

    Billing rules

    Charges are based on the token usage of built-in models. Billing rules vary by model:

    • qwen-plus/qwen3.5-plus/qwen3.6-plus/qwen3.7-max/qwen3.7-plus/qwen3.6-flash/qwen3.8-max/qwen3.7-flash/glm-5.2/deepseek-v4-pro/kimi-k3: Charges are based on both input tokens and output tokens. Input tokens that hit the context cache are charged at the cache hit price.

    • qwen-mt-plus: A translation model. Charges are based on both input tokens and output tokens.

    • qwen3-vl-embedding/tongyi-embedding-vision-plus/text-embedding-v4: Embedding models. Charges are based on input tokens only (no output token fees).

    The fee for model calls is calculated as: token usage within 1 hour × Unit price for the corresponding region, model, and input/output

    Context cache

    For models that support the context cache, input tokens that hit the cache are charged at the cache hit price, which is 20% of the input price in the same region. For cache support and prices of each model, see the price tables below.

    Billing cycle

    Fees are settled hourly (UTC+8). At the end of each billing cycle, the system generates a bill and deducts the fees from your account. Billing data may lag behind actual usage. For details, see Understand Alibaba Cloud bills.

    Pricing

    Important

    For the latest pricing, see the product purchase page.

    Currently, AI Function supports inference with the following built-in models. Each model is billed separately for input and output tokens.

    qwen-plus

    Region

    Token range

    Input price ($/1k tokens)

    Output price ($/1k tokens)

    • China (Beijing)

    • China (Shanghai)

    • China (Hangzhou)

    • China (Shenzhen)

    • China (Qingdao)

    • China (Zhangjiakou)

    • China (Ulanqab)

    0<T≤128K

    0.000138

    0.000344

    128K<T≤256K

    0.000414

    0.003442

    256K<T≤1M

    0.000827

    0.008257

    • Singapore

    • China (Hong Kong)

    • Germany (Frankfurt)

    • US (Virginia)

    • US (Silicon Valley)

    • Japan (Tokyo)

    • Indonesia (Jakarta)

    • Mexico (Queretaro)

    0<T≤256K

    0.000480

    0.001440

    256K<T≤1M

    0.001440

    0.004320

    qwen3.5-plus

    Region

    Token range

    Input price ($/1k tokens)

    Output price ($/1k tokens)

    • China (Beijing)

    • China (Shanghai)

    • China (Hangzhou)

    • China (Shenzhen)

    0<T≤128K

    0.000138

    0.000826

    128K<T≤256K

    0.000344

    0.002064

    256K<T≤1M

    0.000688

    0.004128

    • China (Hong Kong)

    • Singapore

    • Germany (Frankfurt)

    • US (Virginia)

    • US (Silicon Valley)

    • Japan (Tokyo)

    • Indonesia (Jakarta)

    • Mexico (Queretaro)

    0<T≤256K

    0.000480

    0.002880

    256K<T≤1M

    0.000600

    0.003600

    qwen3.6-plus

    Region

    Token range

    Input price ($/1k tokens)

    Output price ($/1k tokens)

    • China (Beijing)

    • China (Shanghai)

    • China (Hangzhou)

    • China (Shenzhen)

    0<T≤256K

    0.000331

    0.001981

    256K<T≤1M

    0.001321

    0.007927

    • China (Hong Kong)

    • Singapore

    • Germany (Frankfurt)

    • US (Virginia)

    • US (Silicon Valley)

    • Japan (Tokyo)

    • Indonesia (Jakarta)

    • Mexico (Queretaro)

    0<T≤256K

    0.000600

    0.003600

    256K<T≤1M

    0.002400

    0.007200

    qwen3.7-max

    qwen3.7-plus

    qwen3.6-flash

    qwen3.8-max

    qwen3.7-flash

    glm-5.2

    deepseek-v4-pro

    kimi-k3

    qwen-mt-plus

    Embedding models

    Model

    Billing type

    Region

    Input price ($/1K tokens)

    qwen3-vl-embedding

    image_embedding

    China mainland regions

    0.0003096

    text-embedding-v4

    context_embedding

    China mainland regions

    0.000086

    text-embedding-v4

    context_embedding

    International regions

    0.000084

    tongyi-embedding-vision-plus

    image_embedding

    International regions

    0.000108

    Billing examples

    The following examples show how AI Function fees are calculated.

    Billing conditions

    Fee (USD)

    • Region: US (Virginia)

    • Function called: ai_complete() (using the qwen-plus model)

    • Input tokens: 750

    • Output tokens: 20

    Fee = Input fee + Output fee

    = 0.000480 × 750/1000 + 0.001440 × 20/1000

    = 0.000360 + 0.000029

    = 0.000389 (USD)