gpt-5.2 vs deepseek-r2
On a 1,500-input, 500-output call at 10,000 calls a day, gpt-5.2 costs about $2,100 a month and deepseek-r2 about $578 — deepseek-r2 is roughly 73% cheaper than gpt-5.2. gpt-5.2 lists $8.00 per 1M output tokens against deepseek-r2 at $2.20.
gpt-5.2 costs $2.00/1M in and $8.00/1M out; deepseek-r2 costs $0.55/1M in and $2.20/1M out.
Last updated . Model pricing is refreshed twice daily.
Price per 1M tokens
gpt-5.2: input $2.00, output $8.00. deepseek-r2: input $0.55, output $2.20.
- Monthly at 10,000 calls/day — gpt-5.2: $2,100, deepseek-r2: $578
- Monthly at 100,000 calls/day — gpt-5.2: $21,000, deepseek-r2: $5,775
- Time to first token — gpt-5.2: 200 ms, deepseek-r2: 300 ms
- Throughput — gpt-5.2: 220 tok/s, deepseek-r2: 45 tok/s
Which one to run
If your workload tolerates deepseek-r2's quality, the saving is immediate. If it does not, route: send the easy calls to deepseek-r2 and keep gpt-5.2 for the hard ones.
Frequently asked questions
Is gpt-5.2 cheaper than deepseek-r2?
deepseek-r2 is cheaper. At 10,000 calls a day with 1,500 input and 500 output tokens, gpt-5.2 costs about $2,100 a month and deepseek-r2 about $578.
What is the price difference between gpt-5.2 and deepseek-r2?
About 73% on the same workload, using list prices with no caching or batch discounts.