DeepSeek V4 Pro 0813

As a 1.7T-parameter frontier model, DeepSeek V4 Pro 0813 works well alongside its smaller sibling, V4 Flash 0731, in coding agents. With Pro powering a main agent and Flash running sub-agents, your whole coding stack runs on open weight models.

DeepSeek V4 Pro offers substantial cost savings versus frontier closed models:

• Input tokens (cache miss): $1.32/m

• Input tokens (cache hit): $0.132/m

• Output tokens: $3.96/m

Whether you’re running evals, writing code, or powering agents in your product, try DeepSeek V4 Pro 0813 on Baseten today.

See HuggingFace model card.