Alibaba Cloud Model Studio:Billing for model services

Tokens are the basic units used by models to represent natural language text, which can be understood as "characters" or "words".

• For Chinese text, 1 token usually corresponds to 1 Chinese character or word. For example, "你好，我是通义千问" will be converted to ['你好', '，', '我是', '通', '义', '千', '问'].

• For English text, 1 token usually corresponds to 3-4 letters or 1 word. For example, "Nice to meet you." will be converted to ['Nice', ' to', ' meet', ' you', '.'].

Different models may have different tokenization methods. You can use the SDK to view the tokenization data of the Qwen model locally.

# Before running: pip install dashscope
from dashscope import get_tokenizer

# Get the tokenizer object, currently only supports the Qwen series models
tokenizer = get_tokenizer('qwen-turbo')

input_str = 'Qwen has powerful capabilities.'

# Split the string into tokens and convert to token ids
tokens = tokenizer.encode(input_str)
print(f"Token IDs after tokenization: {tokens}.")
print(f"There are {len(tokens)} tokens after tokenization")

# Convert token ids back to strings and print them
for i in range(len(tokens)):
    print(f"Token ID {tokens[i]} corresponds to the string: {tokenizer.decode(tokens[i])}")

// Copyright (c) Alibaba, Inc. and its affiliates.
// dashscope SDK version >= 2.13.0
import java.util.List;
import com.alibaba.dashscope.exception.NoSpecialTokenExists;
import com.alibaba.dashscope.exception.UnSupportedSpecialTokenMode;
import com.alibaba.dashscope.tokenizers.Tokenizer;
import com.alibaba.dashscope.tokenizers.TokenizerFactory;

public class Main {
  public static void testEncodeOrdinary(){
    Tokenizer tokenizer = TokenizerFactory.qwen();
    String prompt ="If you were to walk 100,000 miles now, how long would it take to arrive? ";
    // encode string with no special tokens
    List<Integer> ids = tokenizer.encodeOrdinary(prompt);
    System.out.println(ids);
    String decodedString = tokenizer.decode(ids);
    assert decodedString == prompt;
  }

  public static void testEncode() throws NoSpecialTokenExists, UnSupportedSpecialTokenMode{
    Tokenizer tokenizer = TokenizerFactory.qwen();
    String prompt = "<|im_start|>system\nYour are a helpful assistant.<|im_end|>\n<|im_start|>user\nSanFrancisco is a<|im_end|>\n<|im_start|>assistant\n";
    // encode string with special tokens <|im_start|> and <|im_end|>
    List<Integer> ids = tokenizer.encode(prompt, "all");
    String decodedString = tokenizer.decode(ids);
    System.out.println(ids);
    assert decodedString == prompt;

  }

  public static void main(String[] args) {
      try {
        testEncodeOrdinary();
        testEncode();
      } catch (NoSpecialTokenExists | UnSupportedSpecialTokenMode e) {
        e.printStackTrace();
      }
  }
}

You can use this local tokenizer to estimate the token amount of your text, but the result may not be completely consistent with the actual server. If you are interested in the details of the Qwen tokenizer, see Tokenization.

You can check the call count and token consumption for a specific model on the Model Observation page of the console.

In multi-round conversations, the input and output from previous interactions are all billed as new input tokens.

No, you are not. Creating an application alone does not incur charges. You are only billed for model inference if you test or call the application.

If you encounter a balance shortage or overdue payment while using Model Studio, visit Expenses and Costs to pay.

You can set quota alert in the Expenses and Costs center.

You cannot stop pay-as-you-go billing. But as long as you stop using the features of Model Studio, you will not incur fees.

To prevent unexpected API invocation fees, you can delete all your API Key.

Additionally, you can set monthly consumption alert. You will be notified immediately in case of unexpected charges.

1. Go to the Cost Analysis page.

2. Select Pretax Amount for Cost Type.

3. Select Month for Time Unit.

4. Select Alibaba Cloud Model Studio for Product Name.

1. Go to the Cost Analysis page.

2. Select Pretax Amount for Cost Type.

3. Choose a time range.

4. Select Model Studio Foundation Model Inference for Product Detail.

1. Go to the Billing Details tab of the Bill Details page.

2. Select a Billing Cycle.

3. Select Model Studio Foundation Model Inference for Product Details.

4. Click Search. Take qwen-max as an example:

   In the Instance ID column, you can find the input_tokens and output_tokens instances of qwen-max. Add the amounts of the two instances to get the cost of calling the qwen-max model.

Bills generated after September 7, 2024, can be allocated based on: workspace ID, model name, input/output type, and calling channel.

1. Go to the Billing Details tab of the Bill Details page.

2. Select a Billing Cycle.

3. Select Model Studio Foundation Model Inference for Product Details.

4. Click Search.

5. Click Export Billing Overview (CSV) to download the search results.

6. Allocate the costs based on Instance ID.

   The Instance ID, such as text_token;llm-xxx;qwen-max;output_token;app, represents billing type;workspace ID;model name;input/output type;calling channel.

Calling channels include app, bmp, and assistant-api. app refers to model calls through applications, bmp to calls made on the Home or Playground pages of the console, and assistant-api to calls through the Assistant API.

1. Service not activated

Use your Alibaba Cloud account to log on to Expenses and Costs. Activate Model Studio and claim the free quota.

2. Insufficient account balance

• Check balance: Log on to Expenses and Costs. Check whether your account has sufficient balance.

• Recharge: Click Recharge. Enter the desired amount and complete the payment.

3. Set consumption alert to prevent repeated errors

• Set alert: Set monthly consumption alert. The system will send an alert when triggered.