Model comparison
DeepSeek-R1-0528 vs DeepSeek-R1
- Input
- $0.50 / 1M tokens
- Output
- $2.15 / 1M tokens
- Context
- 164K
- Released
- 2025-05-28
- Input
- $0.55 / 1M tokens (cache miss; $0.14 cache hit)
- Output
- $2.19 / 1M tokens
- Context
- 128K
- Released
- 2025-01-20
DeepSeek-R1-0528 scores higher on 7 of the 7 benchmarks both models report; DeepSeek-R1-0528 is cheaper on input tokens ($0.5 vs $0.6 per 1M, 1.1×); DeepSeek-R1-0528 has the larger context window (164K vs 128K).
Specs and pricing side by side
| Spec | DeepSeek-R1-0528 | DeepSeek-R1 |
|---|---|---|
| Maker | DeepSeek | DeepSeeksame |
| Released | 2025-05-28 | 2025-01-20 |
| Context window | 164K | 128K |
| Max output | 32K tokens (64K max generation length) | 32,768 tokens (max generation length) |
| Input price | $0.50 / 1M tokens | $0.55 / 1M tokens (cache miss; $0.14 cache hit) |
| Output price | $2.15 / 1M tokens | $2.19 / 1M tokens |
| Parameters | 671B total / 37B active (685B on Hugging Face incl. MTP module) | 671B total / 37B active (Mixture-of-Experts) |
| License | MIT | MITsame |
| Weights | Open weights | Open weightssame |
| Modalities | Text | Textsame |
| Knowledge cutoff | Not officially disclosed | Not officially disclosedsame |
Benchmarks both models report
| Benchmark | DeepSeek-R1-0528 | DeepSeek-R1 |
|---|---|---|
| AIME 2024 | 91.4% | 79.8% |
| GPQA Diamond | 81% | 71.5% |
| LiveCodeBench | 73.3% | 65.9% |
| Aider-Polyglot | 71.6% | 53.3% |
| MMLU-Pro | 85% | 84% |
| MMLU-Redux | 93.4% | 92.9% |
| FRAMES | 83% | 82.5% |
Scores are as published by each maker or its cited source; harness, reasoning effort and date can differ between rows, so treat small gaps as ties. Sources are listed on each model page.
Which one should you pick?
Pick DeepSeek-R1-0528 if…
- You care about AIME 2024, GPQA Diamond, LiveCodeBench (it scores higher on 7 shared benchmarks).
- Cost matters: input tokens cost 1.1× less ($0.5 vs $0.6 per 1M).
- You need the longer context (164K vs 128K).
- Your use case is: competition-style math and multi-step logical reasoning.
Pick DeepSeek-R1 if…
- Your use case is: competition-style math and multi-step logical reasoning.
FAQ
Which is better, DeepSeek-R1-0528 or DeepSeek-R1?
DeepSeek-R1-0528 scores higher on 7 of the 7 benchmarks both models report; DeepSeek-R1-0528 is cheaper on input tokens ($0.5 vs $0.6 per 1M, 1.1×); DeepSeek-R1-0528 has the larger context window (164K vs 128K). The better pick depends on whether you value benchmark scores, price per token, context length or openness for your workload.
How much cheaper is DeepSeek-R1-0528?
DeepSeek-R1-0528: $0.50 / 1M tokens input, $2.15 / 1M tokens output. DeepSeek-R1: $0.55 / 1M tokens (cache miss; $0.14 cache hit) input, $2.19 / 1M tokens output (list API prices as published by each maker).
What are the context windows of DeepSeek-R1-0528 and DeepSeek-R1?
DeepSeek-R1-0528 offers 164K with up to 32K tokens (64K max generation length) output; DeepSeek-R1 offers 128K with up to 32,768 tokens (max generation length) output.
Is DeepSeek-R1-0528 newer than DeepSeek-R1?
DeepSeek-R1-0528 was released 2025-05-28; DeepSeek-R1 was released 2025-01-20.