DeepSeek: DeepSeek V4 Pro 0813

DeepSeek V4 Pro 0813 is a text model from DeepSeek, identified on api14 as deepseek/deepseek-v4-pro-0813. It is a large-scale mixture-of-experts model, and this build is the GA release of DeepSeek V4 Pro. If you have used earlier DeepSeek models, the family name should be familiar; the 0813 suffix marks this specific release rather than a separate product line.

The model works with text in and text out. It accepts text prompts and returns text, so it does not take images, audio or other media as input. The context window is 1,048,576 tokens, which is large enough to hold entire codebases, long document sets or extended conversation histories in a single request. The maximum output length is 393,216 tokens, so responses can be long as well, not just the inputs. Reasoning is part of the model's surface: the supported parameters include reasoning, reasoning_effort and include_reasoning, which means you can control how much reasoning work the model does and whether that reasoning is returned alongside the final answer.

Those capabilities point to a few practical uses. Long-context work is the obvious fit: analysing large repositories, summarising or querying document collections, and reviewing contracts or specifications that would not fit in a smaller window. Because tool calling is supported, the model can drive agent-style workflows where it decides which functions to call and in what order. Structured outputs and response_format make it suitable for extraction and classification tasks where you need JSON or another schema-conformant shape rather than free prose. The reasoning controls help on tasks that benefit from deliberate multi-step thinking, such as debugging, planning or mathematical work, while still letting you dial that effort down for simpler requests.

You call it through api14's OpenAI-compatible API using the model id deepseek/deepseek-v4-pro-0813. Because the interface follows the OpenAI chat completions shape, existing OpenAI SDKs and clients generally work with only a base URL and key change. The request parameters it supports are frequency_penalty, include_reasoning, logit_bias, logprobs, max_tokens, min_p, presence_penalty, reasoning, reasoning_effort, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs and top_p. In practice that covers sampling controls, determinism via seed, token limits, stop sequences, tool and function calling, structured output, and both log probabilities and reasoning visibility.

One thing to keep in mind is that this is a text-only model. If your input is an image, audio or video, you will need a model that accepts those modalities, since DeepSeek V4 Pro 0813 does not. The large context and output limits also mean requests can be heavy, so for short, simple tasks a smaller model may respond faster and be easier to reason about. Within text work, though, the combination of a very large window, tool calling, structured outputs and reasoning controls makes it a reasonable default for long-document and agentic pipelines.