DeepSeek: DeepSeek V4.1 Flash

DeepSeek V4.1 Flash accepts text and image input and offers a context window of 1,048,576 tokens, with up to 943,718 tokens available for a single completion. That leaves room for very long documents in one request and for unusually long outputs. It supports tool calling and reasoning. Structured output support is not listed, so confirm it before relying on schema-constrained responses. Pricing is $0.051 per million input tokens and $0.567 per million output tokens. Output costs roughly eleven times as much as input, so the economics favor read-heavy work: large document ingestion, image-plus-text analysis, and agent loops that resend long context. Generation-heavy jobs see a smaller saving. There is no independent benchmark coverage yet, so quality is unproven. Shortlist it for cost-sensitive, long-context trials, and test it on your own tasks before committing production traffic.