DeepSeek-R1 · GitHub Models · GitHub

DeepSeek-R1 excels at thinking tasks using a step-by-step training process, such as language, scientific thinking, and coding tasks. It includes 671B total specifications with 37B active specifications, and 128k context length.

DeepSeek-R1 develops on the development of earlier reasoning-focused designs that enhanced efficiency by extending Chain-of-Thought (CoT) thinking. DeepSeek-R1 takes things further by combining support learning (RL) with fine-tuning on carefully chosen datasets. It developed from an earlier version, DeepSeek-R1-Zero, which relied exclusively on RL and revealed strong reasoning abilities however had concerns like hard-to-read outputs and language disparities. To address these constraints, DeepSeek-R1 integrates a percentage of cold-start information and follows a refined training pipeline that mixes reasoning-oriented RL with monitored fine-tuning on curated datasets, leading to a design that accomplishes advanced performance on thinking criteria.

Usage Recommendations

We recommend sticking to the following setups when utilizing the DeepSeek-R1 series designs, including benchmarking, to achieve the expected efficiency:

– Avoid adding a system timely; all directions ought to be included within the user prompt.
– For problems, it is a good idea to include an instruction in your timely such as: « Please reason step by action, and put your last response within boxed . ».
– When evaluating model performance, it is recommended to perform numerous tests and balance the results.

Additional suggestions

The design’s thinking output (consisted of within the tags) might contain more hazardous material than the design’s last action. Consider how your application will utilize or display the thinking output; you might want to reduce the reasoning output in a production setting.