Processing or generating 2,500 words (approximately 3,333 tokens) with DeepSeek V4 Flash ranges from $0.000047 (cached input) to $0.0044 (full generation).
You send 2,500 words as prompt context, documentation, or background knowledge.
DeepSeek V4 Flash drafts a complete 2,500 words article, chapter, or code module.
2,000 words input prompt + 500 words output reply.
| Model | Provider | Input Cost (2,500 words) | Cached Input Cost | Output Cost (2,500 words) | Context Limit |
|---|---|---|---|---|---|
| DeepSeek V4 Flash (Current) | deepseek | $0.001467 | $0.000047 | $0.0044 | 1,000,000 |
| GPT-5.6 Luna | openai | $0.00064 | $0.000064 | $0.00384 | 1,050,000 |
| GPT-5.4 mini | openai | $0.0024 | $0.00024 | $0.0144 | 256,000 |
| GPT-5.4 nano | openai | $0.00064 | $0.000064 | $0.004 | 128,000 |
| GPT-4o mini | openai | $0.00048 | $0.00024 | $0.00192 | 128,000 |
| Claude Haiku 4.5 | anthropic | $0.0033 | $0.00033 | $0.0165 | 1,000,000 |
For DeepSeek V4 Flash (deepseek_bpe tokenizer), 2,500 words is approximately 3,333 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 2,500 words as input costs $0.001467 (or $0.000047 with prompt caching). Generating 2,500 words as output costs $0.0044. 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 $0.000047). For non-urgent asynchronous batch processing, 24-hour batch queues provide a flat 50% discount across all tokens (costing $0.000733 for input, $0.0022 for output).
At an average human reading speed of 250 words per minute, 2,500 words takes approximately 10 minutes to read. In contrast, DeepSeek V4 Flash can process or generate this text in seconds.