Performance on par with OpenAI-o1
Official DeepSeek release of the DeepSeek-R1 language reasoning model, presented as a first-generation reasoning model with open-source weights, MIT licensing, API access via deepseek-reasoner, and a 128K context length.
Public API pricing is token-based. DeepSeek-R1 is exposed as deepseek-reasoner in the API docs; cache-hit input is discounted versus cache-miss input.
DeepSeek's official R1 release page states that DeepSeek-R1 is used via model=deepseek-reasoner and lists public API pricing as $0.14 per 1M input tokens (cache hit), $0.55 per 1M input tokens (cache miss), and $2.19 per 1M output tokens. The official pricing page for deepseek-reasoner matches those figures.
