DeepSeek R1

Model Family

DeepSeek's open reasoning-model family trained with reinforcement learning.

Models in this family

Model	Family	Released	Modality	Context	Open source	API	Status
DeepSeek-R1-0528	DeepSeek R1	May 28, 2025	—	—	Yes	—	active
DeepSeek-R1	DeepSeek R1	Jan 20, 2025	—	—	Yes	—	active
DeepSeek-R1 Distill Llama 70B	DeepSeek R1	Jan 20, 2025	—	—	Yes	—	active
DeepSeek-R1 Distill Llama 8B	DeepSeek R1	Jan 20, 2025	—	—	Yes	—	active
DeepSeek-R1 Distill Qwen 1.5B	DeepSeek R1	Jan 20, 2025	—	—	Yes	—	active
DeepSeek-R1 Distill Qwen 14B	DeepSeek R1	Jan 20, 2025	—	—	Yes	—	active
DeepSeek-R1 Distill Qwen 32B	DeepSeek R1	Jan 20, 2025	—	—	Yes	—	active
DeepSeek-R1 Distill Qwen 7B	DeepSeek R1	Jan 20, 2025	—	—	Yes	—	active
DeepSeek-R1-Zero	DeepSeek R1	Jan 20, 2025	—	—	Yes	—	active

Jan 20, 2025DeepSeek R1 reasoning family launchesDeepSeek R1Research paper

DeepSeek introduced its reinforcement-learning reasoning model family and technical report.