Compare
deepseek-r1
Comparing 1 model
| DeepSeek R1 | |
|---|---|
| Overview | |
| Status | Active |
| Released | 1/20/2025 |
| Summary | Open-weight reasoning model matching o1 performance, fully open-source. |
| Pricing (per 1M tokens) | |
| Input | $0.55/1M |
| Output | $2.19/1M |
| Context and limits | |
| Context window | 128K |
| Max output | 32K |
| Modalities | |
| Supported | text |
| Capabilities | |
| Tool calling | No |
| Structured outputs | No |
| Open source | Yes |
| License | MIT |
| Benchmarks | |
| Aider Polyglot % pass@2 | 56.9 Third-party |
| AIME 2024 % accuracy | 79.8 Self-reported |
| GPQA Diamond % accuracy | 71.5 Self-reported |
| MATH % accuracy | 97.3 Self-reported |
| MMLU % accuracy | 90.8 Self-reported |