DeepSeek-R1
DeepSeek·
LLMsreasoningcloud + localFrontier
Overview
Reasoning model capable of self-verification.
Capabilities and innovations
Chain of thoughtSelf-verificationRL trainingPure RL ReasoningSelf-verificationEmergent CoT Behavior
Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| GPQA Diamond | 71.5% | unsourced |
| Humanity’s Last Exam (no tools) | 18.4% | unsourced |
| SWE-bench Verified | 49.2% | unsourced |
| MMLU | 90.8% | unsourced |
| MMLU-Pro | 83.18% | unsourced |
- SWE-bench Pro: benchmark did not exist at release
- Terminal-Bench 2.x: benchmark did not exist at release
- MMMU-Pro: benchmark did not exist at release
Architecture and hardware
- Parameters
- 671B total, 37B active per token (MoE)
- Estimated VRAM at Q4
- ~386 GB, Frontier class
Links
More from DeepSeek
Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.