DeepSeek: DeepSeek V4 Pro

DeepSeek V4 Pro is a text-in, text-out model from DeepSeek with a context window of roughly one million tokens and a maximum completion length of 384,000 tokens. It supports tool use and reasoning, which makes it applicable to multi-step agentic workflows, though structured output support is unconfirmed. The model is not free and weight availability has not been disclosed. At $0.435 per million input tokens and $0.87 per million output tokens, it sits in a mid-to-lower cost tier for capable models. Its blended benchmark score of 82.5 spans only four benchmarks, so confidence in that figure is limited; notably, its coding subscore of 97.9 is high, while its agentic score of 60.0 is more modest. Teams prioritizing code generation on a constrained budget have a reasonable case to shortlist it, but those needing strong agentic performance or verified structured output should test carefully before committing.