DeepSeek V4 Pro 0813 is DeepSeek-AI’s official DeepSeek-V4-Pro release, described as a reasoning-oriented, text-only model for generation, coding, and agentic tool-use workflows. It is identified as the production release that supersedes the preview model and includes a DSpark speculative-decoding module.
The model is a transformer-based sparse mixture-of-experts system with hybrid compressed sparse attention and heavily compressed attention, and it supports multi-turn messages in OpenAI-compatible format.
| Developer | DeepSeek AI |
|---|---|
| Model ID | deepseek/deepseek-v4-pro-0813 |
| Release Date | August 11, 2026 |
| Context Window | 1,048,576 tokens (~1.0M) |
| Max Output Tokens | 384,000 tokens |
| Modalities | Text → Text |
| Capabilities | reasoning, streaming, structured (JSON schema) output, tool/function calling |
Architecture
Transformer-based sparse mixture-of-experts model with hybrid Compressed Sparse Attention, Heavily Compressed Attention, Manifold-Constrained Hyper-Connections, and an attached DSpark speculative-decoding module.
Scale
The model card lists 1.65T total parameters and 49B active parameters, with a vocabulary size of 129,280.
Input and output
It accepts text input and produces text output as string completions and reasoning content.
Reasoning support
The model card states it supports low, high, and max reasoning-effort levels.
Use profile
It is described as designed for text generation, reasoning, coding, and agentic tool-use workflows.
Reasoning workflows
Suitable for multi-step problem solving and other tasks where the model’s reasoning-effort controls are useful.
Coding assistance
Explicitly described as designed for coding tasks.
Agentic tool use
The model is positioned for agentic workflows and tool-use scenarios.
Long-context analysis
Its stated one-million-token input context makes it suitable for long-form text analysis and multi-turn sessions.
DeepSeek V4 Pro preview
The official release supersedes the preview version and adds DSpark speculative decoding.
DeepSeek V4 Flash
DeepSeek’s API docs list DeepSeek-V4-Flash alongside DeepSeek-V4-Pro as a separate model family entry.
Sources: https://build.nvidia.com/deepseek-ai/deepseek-v4-pro-0813/modelcard · https://fireworks.ai/models/deepseek-ai/deepseek-v4-pro-0813 · https://www.unite.ai/deepseek-ships-v4-pro-as-its-flagship-model-leaves-preview/
Measured on the AI/ML API gateway across live traffic. Figures are medians unless stated. Last updated 2026-10-12.
Text generation, reasoning, coding, and agentic tool-use workflows are documented use cases for DeepSeek-V4-Pro-0813. The model is also described as having greatly enhanced agentic capabilities and being especially strong in production environments.
NO_SOURCE
import requests r = requests.post( "https://api.aimlapi.com/v1/chat/completions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "deepseek/deepseek-v4-pro-0813", "messages": [ { "role": "user", "content": "Hello!" } ] }, ) print(r.json())
const r = await fetch("https://api.aimlapi.com/v1/chat/completions", { method: "POST", headers: { Authorization: `Bearer ${process.env.AIMLAPI_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ "model": "deepseek/deepseek-v4-pro-0813", "messages": [ { "role": "user", "content": "Hello!" } ] }), }); console.log(await r.json());
curl -X POST https://api.aimlapi.com/v1/chat/completions \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"deepseek/deepseek-v4-pro-0813","messages":[{"role":"user","content":"Hello!"}]}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Cached input |
DeepSeek V4 Pro 0813 has documented time-dependent pricing with off-peak rates and higher peak-hour rates, and cached input is billed at a separate discounted rate.
| Benchmark | Score | What it measures | Source | Retrieved |
|---|---|---|---|---|
| Intelligence | 36.3 | Composite score across standardised reasoning, knowledge and problem-solving evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Coding | 68.8 | Composite score across standardised coding evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
DeepSeek V4 Pro 0813 This page | Agentic workflows and structured output |
DeepSeek V4 Pro 0813 has a 1,048,576 tokens context window and can return up to 384,000 tokens.
DeepSeek V4 Pro 0813 takes text as input and returns text.
Use deepseek/deepseek-v4-pro-0813 as the model id. Requests go to https://api.aimlapi.com/v1/chat/completions.
DeepSeek V4 Pro 0813 became available on August 12, 2026.
DeepSeek V4 Pro 0813 is priced at input $2.393196 / 1M tokens, output $4.786392 / 1M tokens, cached input $0.199433 / 1M tokens.
Yes, DeepSeek V4 Pro 0813 can stream responses as they are generated.
DeepSeek V4 Pro 0813 was built by DeepSeek.
