DeepSeekDeepSeek R1 0528
DeepSeek R1 0528 is a DeepSeek open-weights reasoning model, released 28 May 2025. Its best configuration ranks #313 of 372 on the BenchLeader Index at 49.7 ±6.9, in the lower half. It scores highest in human preference (61) and lowest in knowledge (35). At $1.00 per million tokens blended it is mid-priced. Output speed of 24 tokens per second puts it in the slowest quarter, with a first token in 1.5 s. Last measured 2 Sept 2026.
- Blended price
- $1.00/M
- $0.574 in · $2.29 out
- Output speed
- 24 tok/s
- OpenRouter traffic, 7-day median; not yet measured by Artificial Analysis
- First answer
- 1.52 s
- Context
- 164k
- Overall index50
- Reasoning52
- Coding53
- Maths48
- Knowledge35
- Human preference61
Versions
DeepSeek has shipped 2 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.
| Model | Released | Index | Rank |
|---|---|---|---|
| DeepSeek R1 0528this page | 28 May 2025 | 49.7 | #313 |
| DeepSeek R1 | 20 Jan 2025 | 49.7 | #312 |
Benchmark results
One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | default | Source |
|---|---|---|
| GPQA Diamond | 76.3%#131 | Epoch AI Benchmarking Hub |
| SimpleBench | 40.8%#60 | SimpleBench |
| LMArena Hard Prompts | 1433#113 | LMArena |
Coding
| Benchmark | default | Source |
|---|---|---|
| WeirdML | 41.6%#96 | WeirdML |
| LMArena Coding | 1464#108 | LMArena |
Maths
| Benchmark | default | Source |
|---|---|---|
| OTIS Mock AIME | 66.4%#139 | Epoch AI Benchmarking Hub |
| MATH Level 5 | 96.6%#9 | Epoch AI Benchmarking Hub |
| IMO 2025 | 6.8%#6 | MathArena |
| MathArena Apex | 1.0%#35 | MathArena |
Knowledge
| Benchmark | default | Source |
|---|---|---|
| PRBench Finance | 32.7%#31 | Scale AI SEAL |
| PRBench Legal | 36.6%#28 | Scale AI SEAL |
| MultiNRC | 27.6%#27 | Scale AI SEAL |
Human preference
| Benchmark | default | Source |
|---|---|---|
| LMArena Text | 1421#105 | LMArena |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| StreamLake | 52 tok/s | 3.28 s | $0.571 | $2.29 | 128k | – |
| SiliconFlow | 24 tok/s | 1.52 s | $0.500 | $2.18 | 164k | fp8 |
| DeepInfra | 17 tok/s | 0.86 s | $0.500 | $2.15 | 164k | fp4 |
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.350 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0009 | $0.0009 | 14.0 s |
| Summarise a 30-page report | 12,000 / 600 | $0.0083 | $0.0062 | 26.5 s |
| Code edit | 6,000 / 1,500 | $0.0069 | $0.0059 | 1.1 min |
| Agentic coding session | 60,000 / 4,000 | $0.044 | $0.034 | 2.8 min |
| Structured extraction | 2,000 / 200 | $0.0016 | $0.0013 | 9.9 s |
See also
Data as of 10 Sept 2026. Compare with another model.