DeepSeek V4 Pro API

DeepSeek V4 Flash and Pro are large language models for chat and reasoning tasks with up to 1M context length. They support both standard and thinking modes, enabling efficient responses or deeper reasoning workflows.

Modalities
TextText
In / Out price
$0.9/$2.7per 1M
Context window
1M tokens
Released
Apr 24, 2026

DeepSeek V4 Pro overview

DeepSeek V4 Pro specs

Model ID
deepseek/deepseek-v4-pro
Developer
DeepSeek
Context window
1,000,000 tokens
Max output
384,000 tokens
Input
Text
Output
Text
API
OpenAI-compatible, base URL https://api.aimlapi.com/v1

What Is DeepSeek V4 Pro?

DeepSeek V4 Pro is the flagship model from DeepSeek's fourth-generation release, dropped on April 24, 2026. It is the largest open-weights model currently available, larger than Kimi K2.6 at 1.1T and more than twice the size of its predecessor, DeepSeek V3.2 at 685B. That scale alone would mean little without efficiency; what makes V4 Pro genuinely remarkable is how little of that scale it uses during inference.

Using a Mixture-of-Experts (MoE) design, V4 Pro activates only 49 billion parameters per token, roughly 3% of its full weight. In the one-million-token context setting, it requires just 27% of the inference FLOPs and 10% of the KV cache size compared with DeepSeek V3.2. Those are not incremental improvements. They represent a step-change in what's economically feasible to run at production scale.

Three Innovations Behind the Efficiency

Most models label million-token context windows as a marketing feature. At that scale, standard attention is quadratically expensive, memory balloons, inference slows, and costs multiply. DeepSeek solved this with three architectural breakthroughs developed and published before the V4 launch.

Hybrid Attention Architecture

Compressed Sparse Attention (CSA) combined with Heavily Compressed Attention (HCA) replaces standard full attention. The result: 27% of the inference FLOPs and just 10% of the KV cache at 1M tokens, making long-context inference genuinely deployable at scale.

Manifold-Constrained Hyper-Connections

Standard Hyper-Connections caused 3,000× signal amplification in 27B experiments, crashing training. The mHC framework constrains mixing matrices using the Sinkhorn-Knopp algorithm, cutting amplification to 1.6×, enabling stable training at 1.6T parameters.

Muon Optimizer

Pre-training uses the Muon optimizer for faster convergence and training stability versus standard AdamW. At 1.6T-parameter scale, gradient collapse compounds quickly, Muon alongside mHC's stability guarantees made 33T-token training achievable.

Two-Stage Post-Training

Independent SFT and RL cultivation of domain-specific experts (using GRPO), followed by unified model consolidation via on-policy distillation. Each domain's strength is preserved, then blended into a single generalist model without capability regression.

Benchmark Results

DeepSeek benchmarks V4 Pro as competitive with top closed-source models across reasoning, coding, and knowledge tasks. On SWE-bench Verified, a real-world software engineering benchmark, it scores 80.6%, sitting within 0.2 points of Claude Opus 4.6 at roughly one-seventh the output cost.

Reasoning Effort Modes

V4 Pro and V4 Flash both support three configurable reasoning modes, letting you trade off speed against depth depending on what the task actually requires, rather than paying for maximum thinking on every call.

Standard (Non-Thinking)

Default mode. Fast, direct responses without extended chain-of-thought. Best for retrieval, summarization, structured outputs, and tasks where latency matters more than deep multi-step reasoning.

Think

Activates step-by-step reasoning before the final answer. The model works through the problem internally before responding. Visible reasoning tokens appear in the reasoning_details response field. Suitable for complex coding, math, and analytical tasks.

Who Should Use DeepSeek V4 Pro?

V4 Pro's combination of 1M-token context, strong agentic coding performance, and competitive pricing makes it suited to a specific class of workloads. Here is where it fits best, and where you might opt for V4 Flash instead.

Full-Codebase Analysis

At 1M tokens, you can load an entire medium-sized repository into context. V4 Pro's SOTA performance on Terminal-Bench and SWE-bench makes it genuinely capable at cross-file refactoring, bug investigation, and architectural review without truncation.

Long-Horizon Agentic Tasks

Multi-step automation, research synthesis, and complex workflow execution where the agent must track state across many turns. V4 Pro leads open-source models on agentic coding benchmarks and holds comparable performance to V4 Flash on simpler agent tasks.

Math, STEM & Scientific Reasoning

Beats all current open-weight models on math and STEM benchmarks. Competitive with top closed-source models on GPQA Diamond. Suitable for technical research assistance, problem solving, and educational tooling requiring deep domain knowledge.

Knowledge-Intensive Retrieval

V4 Pro ranks first among open models for world knowledge, trailing only Gemini 3.1 Pro overall. Enterprises building RAG pipelines or document-heavy Q&A systems that need strong factual grounding will find V4 Pro's recall noticeably above peer open-source models.

Use cases

Complex problem-solving and planning are documented use cases for DeepSeek V4 Pro, with the model described as suited to “complex problem-solving” and “planning” in its reasoning modes.

Coding and software engineering are also documented use cases, including “complex software engineering,” “coding,” and “complex coding and debugging workflows.”

Agentic workflows are a stated fit for the model, including long-running agentic tasks, tool use scenarios, and agentic workflows requiring planning and decision-making.

Long-document understanding and synthesis is another documented use case, with the model described as suited to “long-document understanding and synthesis” and deeper context understanding.

Capabilities

  • Function calling
  • Streaming
  • Structured outputs
  • Reasoning
  • Web search

DeepSeek V4 Pro benchmarks by Artificial Analysis

See how DeepSeek V4 Pro compares in intelligence, generation speed, cost per task and latency. Data retrieved Nov 3, 2026.

Intelligence

Index v5.0

Speed

Output tokens / second

Cost per task

USD / benchmark task

Benchmark results

Compare 7 models available on AI/ML API across 5 independent benchmarks.

Humanity's Last Exam

Expert-written questions across academic subjects.

DeepSeek V4 Pro API pricing

DeepSeek V4 Pro pricing in US dollars
UsagePriceUnit
Input$0.91M tokens
Output$2.71M tokens
Cached input$0.031M tokens

Prices from the AI/ML API price list, checked on Nov 4, 2026.

DeepSeek V4 Pro vs other models: token prices

AI/ML API input and output rates for 5 models, per million tokens.

AI/ML API model prices in USD per million tokens
ModelInputOutput
DeepSeek V4 Pro$0.9$2.7
DeepSeek Chat (V4.1 Flash)$0.3$1.2
DeepSeek V4.1 Flash$0.15$0.6
DeepSeek V4 Flash Vision Exp$0.44$1.32
DeepSeek V4.1 Flash 1017$0.15$0.6

How to use the DeepSeek V4 Pro API

View docs

Build with your agent

Give your agent the model details, code and documentation.

1Get your API key

Create API key
export AIMLAPI_KEY="8f2b4c7d9a1e6b3f…"

2Make your first request

curl 'https://api.aimlapi.com/v1/chat/completions' \
  -H "Authorization: Bearer $AIMLAPI_KEY" \
  -H 'Content-Type: application/json' \
  -d '{
  "model": "deepseek/deepseek-v4-pro",
  "messages": [
    {
      "role": "user",
      "content": "Write a haiku about the ocean."
    }
  ]
}'

Related articles

All posts

DeepSeek V4 Pro FAQ

What is the context window of DeepSeek V4 Pro?

DeepSeek V4 Pro has a 1,000,000 tokens context window and can return up to 384,000 tokens.

What inputs and outputs does DeepSeek V4 Pro support?

DeepSeek V4 Pro takes text as input and returns text.

What is the model ID for DeepSeek V4 Pro?

Use deepseek/deepseek-v4-pro as the model id. Requests go to https://api.aimlapi.com/v1/chat/completions.

When was DeepSeek V4 Pro released?

DeepSeek V4 Pro became available on April 24, 2026.

How much does DeepSeek V4 Pro cost?

DeepSeek V4 Pro costs $0.9 input and $2.7 output per 1M tokens on AI/ML API, billed as you go with no subscription.

Does DeepSeek V4 Pro support streaming?

Yes, DeepSeek V4 Pro can stream responses as they are generated.

Who created DeepSeek V4 Pro?

DeepSeek V4 Pro was built by DeepSeek.

Start building with DeepSeek V4 Pro

DeepSeek V4 Pro API

deepseek/deepseek-v4-pro

Build with your agent

Give your agent the model details, code and documentation.

1Get your API key

Create API key
export AIMLAPI_KEY="8f2b4c7d9a1e6b3f…"

2Make your first request

curl 'https://api.aimlapi.com/v1/chat/completions' \
  -H "Authorization: Bearer $AIMLAPI_KEY" \
  -H 'Content-Type: application/json' \
  -d '{
  "model": "deepseek/deepseek-v4-pro",
  "messages": [
    {
      "role": "user",
      "content": "Write a haiku about the ocean."
    }
  ]
}'