Head to head
deepseek-v4-pro vs gpt-5-mini: API Pricing Compared
On a typical chat turn of 10K input and 1K output tokens, gpt-5-mini costs $0.00450 against $0.00858 for deepseek-v4-pro, making it 48% cheaper. Both rates are read daily from the providers' own pricing pages.
Side by side
| deepseek-v4-pro | gpt-5-mini | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Input / 1M tokens | $0.66 | $0.25 |
| Output / 1M tokens | $1.98 | $2.00 |
| Context window | 1M | short |
| Blended cost / chat turn | $0.00858 | $0.00450 |
| Price verified | 2026-08-17 | 2026-08-06 |
Cost by workload
The headline rate rarely decides the bill. These are the same four workloads priced against both models.
| Workload | Tokens | deepseek-v4-pro | gpt-5-mini |
|---|---|---|---|
| Single short call | 2K in, 500 out | $0.00231 | $0.00150 |
| Typical chat turn | 10K in, 1K out | $0.00858 | $0.00450 |
| Long document pass | 100K in, 2K out | $0.070 | $0.029 |
| 1,000 chat turns | 10M in, 1M out | $8.58 | $4.50 |
Which one actually wins
gpt-5-mini in practice. deepseek-v4-pro does have the cheaper output rate, so a crossover exists on paper, but it only arrives once the workload emits 20.5 output tokens for every input token. Summarisation sits near 0.05 and even a generation-heavy agent loop rarely passes 2, so that line is not reachable by a real workload and gpt-5-mini is cheaper for essentially all of them. Price is only one axis, so weigh it against context window, latency, and the quality difference on your own evaluation set before switching.
Common questions
Which is cheaper, deepseek-v4-pro or gpt-5-mini?
gpt-5-mini is cheaper for a typical chat turn of 10K input and 1K output tokens, at $0.00450 against $0.00858 for deepseek-v4-pro, a difference of 48%.
How much does deepseek-v4-pro cost per 1M tokens?
DeepSeek charges $0.66 per 1M input tokens and $1.98 per 1M output tokens for deepseek-v4-pro, verified 2026-08-17.
How much does gpt-5-mini cost per 1M tokens?
OpenAI charges $0.25 per 1M input tokens and $2.00 per 1M output tokens for gpt-5-mini, verified 2026-08-06.
Does gpt-5-mini stay cheaper for every workload?
In practice yes. deepseek-v4-pro has the cheaper output rate, but it only overtakes gpt-5-mini once output runs to 20.5 tokens for every input token. Even generation-heavy agent workloads rarely pass 2, so gpt-5-mini is the cheaper choice for effectively any real workload.
Full pricing pages
Maintained by Jordan Hale, TechFastForward. Both prices are read daily from the official provider pages and recorded only when they move. Compare any other pair on the LLM API Pricing Tracker.