Service, Credit & Pricing

Model Services

Model services on GPTBots are provided directly by GPTBots. Developers can use services from platforms like OpenAI and Claude through GPTBots without having to register their own keys on these platforms. Credits are deducted according to the pricing below when model services are called.

Service Credit Pricing

All services within GPTBots are priced and usage is tracked using "credits". Different LLM versions consume different amounts of credits. For detailed consumption calculations, please refer to the following sections.

Note: Credits cannot be refunded or exchanged.

Credit Consumption Types

The GPTBots platform differentiates pricing based on service types (see Service, Credit & Pricing for details), with credits deducted according to different service rates. There are 10 specific billing types. When using AI Agents and Workflows, different types of services will consume corresponding credits. Developers can view credit consumption statistics under "Organization - Usage".

LLM Service Pricing

Note: The following prices are measured in "credits / 1K Tokens".

Embedding Service Pricing

Note: The following prices are measured in "credits / 1K Tokens".

Rerank Service Pricing

Note: The following prices are measured in "credits / 1K Tokens".

ASR Service Pricing

Note: The following prices are measured in "credits / 60 secs".

TTS Service Pricing

Note: The following prices are measured in "credits / 1000 chars".

Content Moderation Service Pricing

Note: The pricing unit below is measured in credits per request.

Vector Storage

Note: The following prices are measured in "credits / 1K Tokens/ day".

FAQ

How to Convert Between GPTBots Credits and Tokens?

Taking OpenAI's LLM service GPT-4.1-1M as an example, inputting 1000 tokens consumes 0.22 credits.
$10 = 1000 credits = 4,545,454 Tokens (1000 credits / 0.22 credits * 1000 tokens)

Note:
These are approximate estimates and actual values may vary depending on text content and tokenization method.
Word counts are easier to estimate for English and other Latin-based languages, while character counts are more relevant for Chinese, Japanese, Korean, Thai, etc.

How Are Tokens Calculated?

Taking OpenAI's LLM service token calculation rules as an example:

1. English: 1 English word ≈ 1.3 tokens, 1 token ≈ 4 English characters (including spaces and punctuation)
2. Chinese: 1 Chinese character ≈ 1 token (sometimes 1.5 tokens, averaged)
3. Japanese: 1 token ≈ 1 Japanese kana/kanji
4. Korean: 1 token ≈ 1 Korean letter (syllable blocks may be longer)
5. French: 1 French word ≈ 1.2 tokens
6. German: 1 German word ≈ 1.2 tokens
7. Thai: 1 token ≈ 1 Thai letter (Thai has no spaces, token count may be higher after tokenization)
8. Russian: 1 Russian word ≈ 1.2 tokens
9. Arabic: 1 Arabic word ≈ 1.2 tokens

   For specific token counting needs, you can use OpenAI's tiktoken tool for actual testing.

How Are Tokens Calculated for Image Inputs?

Taking OpenAI's LLM service token calculation rules as an example, here's how tokens are calculated for images:

1. Get the image's length and width in "px", e.g., "1024px * 1024px".
2. Calculate the image's "Tiles" value by dividing both "width" and "height" by 512, rounding up, and multiplying the results.
3. Calculate the image's "Tokens" using the formula "85+170*Tiles".

• Complete calculation formula:

  $$Tiles = ⌈(width÷512)⌉×⌈(height÷512)⌉$$
  $$Tokens = 85+170×Tiles$$

• Python code example:

                      
                      import math

def calculate_tokens(width, height):
    tiles = math.ceil(width/512) * math.ceil(height/512)
    tokens = 85 + 170 * tiles
    return tokens

# Test
print(calculate_tokens(2000, 500))

                    

For example, if the input image dimensions are 2000px * 500px, its Tiles value would be 4*1=4, so the input Tokens for this image would be 85+170*4=765.