Processing or generating 25,000 words (approximately 33,325 tokens) with Microsoft Phi-3.5 MoE ranges from $0.00125 (cached input) to $0.02 (full generation).
You send 25,000 words as prompt context, documentation, or background knowledge.
Microsoft Phi-3.5 MoE drafts a complete 25,000 words article, chapter, or code module.
20,000 words input prompt + 5,000 words output reply.
| Model | Provider | Input Cost (25,000 words) | Cached Input Cost | Output Cost (25,000 words) | Context Limit |
|---|---|---|---|---|---|
| Microsoft Phi-3.5 MoE (Current) | microsoft | $0.004999 | $0.00125 | $0.02 | 128,000 |
| GPT-5.6 Luna | openai | $0.0064 | $0.00064 | $0.0384 | 1,050,000 |
| GPT-5.4 mini | openai | $0.024 | $0.0024 | $0.144 | 256,000 |
| GPT-5.4 nano | openai | $0.0064 | $0.00064 | $0.04 | 128,000 |
| GPT-4o mini | openai | $0.0048 | $0.0024 | $0.0192 | 128,000 |
| Claude Haiku 4.5 | anthropic | $0.033 | $0.0033 | $0.165 | 1,000,000 |
For Microsoft Phi-3.5 MoE (cl100k_base tokenizer), 25,000 words is approximately 33,325 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 25,000 words as input costs $0.004999 (or $0.00125 with prompt caching). Generating 25,000 words as output costs $0.02. 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 $0.00125). For non-urgent asynchronous batch processing, 24-hour batch queues provide a flat 50% discount across all tokens (costing $0.002499 for input, $0.009998 for output).
At an average human reading speed of 250 words per minute, 25,000 words takes approximately 100 minutes to read. In contrast, Microsoft Phi-3.5 MoE can process or generate this text in seconds.