DeepSeek launches open-source AI model R1 claiming to outperform competitors
sea.mashable.com — January 22, 2025, 03:01 PM UTC
Chinese AI company DeepSeek has launched an open-source large language model called DeepSeek R1. This model claims to outperform popular models like OpenAI's in key areas, particularly in math, coding, and reasoning tasks. DeepSeek R1 improves upon its predecessor, DeepSeek R1 Zero, by addressing issues with readability and language mixing. It uses a new training method that combines multi-stage training and reinforcement learning. The model is free to use as a web app, with low-cost API access. DeepSeek R1 demonstrates strong capabilities, producing effective code and providing useful advice in various tasks.