DeepSeek-R1-0528 is a reinforcement learning (RL) driven inference model that solves the repeatability and readability issues in the model. Before RL, DeepSeek-R1 introduced cold start data to further optimize inference performance. It performs on par with OpenAI-o1 in math, coding, and reasoning tasks, and improves overall performance through carefully designed training methods.
Provider: deepseek-ai
Context window: 160K
Pricing: $0.35 input / $1.50 output per 1M tokens
Capabilities: serverless, json_mode, function_calling, structured_output
