DeepSeek-R1
The historic foundation model that revolutionized reinforcement learning for LLMs. Trained with Group Relative Policy Optimization (GRPO) without a critic model, emitting visible, verifiable Chain-of-Thought (CoT) reasoning tokens before producing definitive answers.
- Not listed on the current DeepSeek API pricing page (deepseek-flash and deepseek-v4-pro only). Use chat.deepseek.com or historical announcements for access.
- 128K Tokens
- 671B Total; active 37B Active / Token
- DeepSeekMoE + Pure RL via Group Relative Policy Optimization (GRPO)
Not listed on the current DeepSeek API pricing page (deepseek-flash and deepseek-v4-pro only). Use chat.deepseek.com or historical announcements for access.
Sourced benchmarks
- MATH-500: 97.3% — Rigorous 500-question subset of high school competition mathematics. Source
- AIME 2024: 79.8% — First open model to achieve parity with proprietary frontier reasoning engines. Source
- Codeforces Rating: 96.3 %tile (2,029 Elo) — Simulated contest performance against global human participants. Source
- MMLU (Overall): 90.8% — 57 multi-discipline subjects covering humanities, STEM, and law. Source
All models · API pricing · Off-peak pricing guide · LLM API price comparison · Model names · FAQ