Processing or generating 500,000 words (approximately 666,500 tokens) with Text Embedding 3 (Small) ranges from $0.003333 (cached input) to $0.00 (full generation).
Quick answer
For 500,000 words, Text Embedding 3 (Small) processes approximately 666,500 tokens. That costs $0.0133 as uncached input, $0.003333 with the modeled cache discount, or $0.00 when generated as output.
Method & trust
The estimate converts the selected English word count with the model tokenizer family, then applies the published input/output price per million tokens. Real prompts vary with code, formatting, language, and system instructions.
You send 500,000 words as prompt context, documentation, or background knowledge.
Text Embedding 3 (Small) drafts a complete 500,000 words article, chapter, or code module.
400,000 words input prompt + 100,000 words output reply.
| Model | Provider | Input Cost (500,000 words) | Cached Input Cost | Output Cost (500,000 words) | Context Limit |
|---|---|---|---|---|---|
| Text Embedding 3 (Small) (Current) | openai | $0.0133 | $0.003333 | $0.00 | 8,191 |
| Text Embedding 3 (Large) | openai | $0.0866 | $0.0217 | $0.00 | 8,191 |
| Text Embedding 004 | $0.0133 | $0.003333 | $0.00 | 8,192 |
For Text Embedding 3 (Small) (cl100k_base tokenizer), 500,000 words is approximately 666,500 tokens (an average ratio of 1.33 tokens per word in English). Code, technical vocabulary, and non-English scripts will have higher token densities.
Sending 500,000 words as input costs $0.0133 (or $0.003333 with prompt caching). Generating 500,000 words as output costs $0.00. Output tokens are more expensive because autoregressive token generation requires significantly more computation per token.
Yes! If your input text is part of a repeated context, prompt caching saves 50% to 90% (costing $0.003333). For non-urgent asynchronous batch processing, 24-hour batch queues provide a flat 50% discount across all tokens (costing $0.006665 for input, $0.00 for output).
At an average human reading speed of 250 words per minute, 500,000 words takes approximately 2000 minutes to read. In contrast, Text Embedding 3 (Small) can process or generate this text in seconds.