DeepSeek released DeepInfer, a high-performance large-scale inference framework for its own model line, on 20 October 2026.

The company first launched the R1 large language model in January 2025. That model was reported to demonstrate reasoning capability equivalent to OpenAI ChatGPT at significantly lower development cost.

Separate reporting includes allegations published by The Information that DeepSeek circumvented United States high-end chip export bans to obtain Nvidia processors for model training.

No public filing states what hardware DeepInfer is optimized to run on.