Processing or generating 10,000 words (approximately 12,800 tokens) with GPT-5.4 Workhorse ranges from $0.0032 (cached input) to $0.192 (full generation).
You send 10,000 words as prompt context, documentation, or background knowledge.
GPT-5.4 Workhorse drafts a complete 10,000 words article, chapter, or code module.
8,000 words input prompt + 2,000 words output reply.
| Model | Provider | Input Cost (10,000 words) | Cached Input Cost | Output Cost (10,000 words) | Context Limit |
|---|---|---|---|---|---|
| GPT-5.4 Workhorse (Current) | openai | $0.032 | $0.0032 | $0.192 | 256,000 |
| GPT-5.6 Terra | openai | $0.0256 | $0.00256 | $0.1536 | 1,050,000 |
| o3-mini | openai | $0.0141 | $0.00704 | $0.0563 | 200,000 |
| o4-mini | openai | $0.0141 | $0.00352 | $0.0563 | 256,000 |
| o1-mini | openai | $0.0141 | $0.00704 | $0.0563 | 128,000 |
| Claude Sonnet 5 | anthropic | $0.0264 | $0.00264 | $0.132 | 1,000,000 |
For GPT-5.4 Workhorse (o200k_base tokenizer), 10,000 words is approximately 12,800 tokens (an average ratio of 1.28 tokens per word in English). Code, technical vocabulary, and non-English scripts will have higher token densities.
Sending 10,000 words as input costs $0.032 (or $0.0032 with prompt caching). Generating 10,000 words as output costs $0.192. 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 $0.0032). For non-urgent asynchronous batch processing, 24-hour batch queues provide a flat 50% discount across all tokens (costing $0.016 for input, $0.096 for output).
At an average human reading speed of 250 words per minute, 10,000 words takes approximately 40 minutes to read. In contrast, GPT-5.4 Workhorse can process or generate this text in seconds.