DeepSeek V3.2 vs GPT-5.6 Luna
DeepSeek against OpenAI, costed on the same three workloads. Neither is cheaper everywhere — which one wins depends on how much you read versus write.
Monthly bill, side by side
| Workload | DeepSeek V3.2 | GPT-5.6 Luna | Difference |
|---|---|---|---|
| Support chatbot 2,500 six-turn conversations a day |
$723 | $887 | $164 (+23%) |
| Document summariser 1,000 single-shot summaries a day |
$112 | $102 | $10.08 (+10%) |
| Coding agent 200 twelve-turn sessions a day |
$606 | $598 | $8.64 (+1%) |
No prompt caching and no batch discount on either side, so the comparison is like for like. Both figures assume 30 days a month.
The rate cards
| DeepSeek V3.2 | GPT-5.6 Luna | |
|---|---|---|
| Input, per 1M tokens | $0.28 | $0.2 |
| Output, per 1M tokens | $0.4 | $1.2 |
| Cached input | not published | $0.02 |
| Context window | 164K | 1.05M |
| Maximum output | 164K | 128K |
| Reasoning model | yes | yes |
| Vision | no | yes |
Run this comparison on your own numbers
Full detail: DeepSeek V3.2 pricing · GPT-5.6 Luna pricing