DeepSeek V4 Pro

Name: DeepSeek V4 Pro
Brand: DeepSeek
SKU: deepseek-v4-pro
Price: 1.584 USD
Availability: InStock

Text & chatNewAvailable

by DeepSeekdeepseek-v4-pro

DeepSeek's April 2026 flagship. 1.6T MoE / 49B active params, 1M context, rivals top closed-source models on STEM and coding at a fraction of the price.

Try it Get API key Compare

Price · 1M in / out: $1.584 / $4.752
Context: 1,048,576 tokens
Max. output: 384,000 tokens
Input → output: Text → Text
Developer: DeepSeek
Updated: September 23, 2026

Playground

Try DeepSeek V4 Pro

Chat

$1.584/1M in

Send a message. The answer arrives in full once the model is done (no streaming).

System prompt

Max. answer length (tokens)

TemperatureModel default

⌘/Ctrl + Enter

This run

at most $0.0049 · 0.49 credits reserved

Billed by the tokens actually used; the unused part of the reservation is refunded.

New here?

10 free credits ($0.10) when you sign up with Google

Usable 24 hours after sign-up, up to 5 runs per day and at most 2 credits per run. Other sign-in methods start without credits. Enough for 20 runs of this model.

About DeepSeek V4 Pro

TL;DRAs of September 23, 2026

DeepSeek V4 Pro is a model by DeepSeek in the Text & chat category. On Railwail, DeepSeek V4 Pro costs $1.584 per 1M input tokens and $4.752 per 1M output tokens. The context window holds 1,048,576 tokens, and one response can be up to 384,000 tokens long.

Released April 24, 2026 as part of the DeepSeek V4 Preview, DeepSeek-V4-Pro is a 1.6T-parameter Mixture-of-Experts model with 49B activated parameters. Native 1M-token context, 384K max output. Tops open-weights leaderboards on Math/STEM/Coding (~81% SWE-bench Verified) and rivals frontier closed-source models. Best for: open-weights coding agents, long-document analysis, cost-efficient reasoning workloads.

Background

About DeepSeek AI

Founded 2023 · Hangzhou, China

DeepSeek AI is a Chinese AI research lab founded in 2023 by Liang Wenfeng, the founder of the High-Flyer quantitative hedge fund. The lab is funded primarily by High-Flyer's profits and operates independently of major Chinese tech conglomerates. DeepSeek's mission is to build open frontier AI: every flagship model has been released with open weights and a permissive license. Major releases include DeepSeek LLM (late 2023), DeepSeek-V2 (May 2024, MoE), DeepSeek-V3 (December 2024, 671B MoE / 37B active), DeepSeek-R1 (January 2026 family, reasoning), DeepSeek V3.1 (early 2026) and the DeepSeek V4 family released as preview April 24, 2026. DeepSeek's models repeatedly top open-weights leaderboards on coding, math and reasoning at a fraction of the training cost claimed by Western labs, and the team is credited with popularising large-scale Reinforcement Learning from Verifiable Rewards.

Visit DeepSeek AI

Architecture

Sparse Mixture-of-Experts Transformer (frontier open-weights)

DeepSeek-V4-Pro was released April 24, 2026 as the flagship of the V4 family. It is a Sparse MoE Transformer with 1.6T total parameters and 49B activated per token, supporting a native 1M-token context window with up to 384K-token max output. The model was trained on the lab's expanded GPU cluster using DeepSeek's signature recipe: large-scale pretraining on a multi-trillion-token mixture of web text, code, books, scientific papers and curated math/STEM data, followed by extensive Reinforcement Learning from Verifiable Rewards (RLVR) on math, coding and tool-use trajectories. Architectural innovations introduced in V3 - Multi-head Latent Attention (MLA), DeepSeekMoE with fine-grained expert specialization and shared experts, and FP8 mixed-precision training - are retained and refined. V4 Pro is published with open weights under a permissive license and runs natively in both server and inference frameworks such as vLLM and SGLang. DeepSeek treats the V4 launch as a preview phase and has announced that the older deepseek-chat and deepseek-reasoner endpoints will be deprecated on July 24, 2026.

Parameters: 1.6T total / 49B active per token
Context: 1,048,576 tokens

Capabilities

1M token native context window with 384K max output
~81% SWE-bench Verified - rivals top closed-source models
Top open-weights scores on Math/STEM/Coding benchmarks
1.6T MoE / 49B active parameters
FP8-trained for compute efficiency
Multi-head Latent Attention for memory-efficient long context
Function calling and structured JSON output
Cache hit pricing at $0.0145 per million tokens enables cheap multi-turn agents
Available via DeepSeek API, OpenRouter, Together, Fireworks and self-hosted with vLLM/SGLang
Open weights under a permissive license
Best for: open-weights coding agents, long-document analysis, cost-efficient reasoning workloads, on-premise enterprise deployments.

Training & license

Pretrained on a multi-trillion-token mixture of web text, code, books, scientific papers and curated math/STEM data. Post-training applies large-scale Reinforcement Learning from Verifiable Rewards (RLVR) on math, coding and tool-use tasks, plus supervised fine-tuning and instruction tuning. Knowledge cutoff approximately early 2026.

License: Open weights under a permissive license that allows commercial use. Hosted API access via deepseek.com.

Safety testing: DeepSeek publishes model cards but provides limited external red-teaming. Safety filters are lighter than Western frontier labs; deployers are responsible for downstream alignment.

Known limitations

Light built-in safety alignment relative to Western frontier models
No native vision or audio input (text-only)
Pro variant requires substantial GPU resources to self-host
Older deepseek-chat / deepseek-reasoner endpoints will be deprecated July 24, 2026
Some Chinese-language safety constraints apply

Research papers

Pricing

Prices in US dollars. Usage is charged from prepaid credits.

Input	$1.584 / 1M tokens
Output	$4.752 / 1M tokens

Billed by the tokens each request actually uses.
1 credit = $0.01

Cost calculator

Price calculator

Input tokens

/ req.

Output tokens

/ req.

Requests

Total

$0.40

40 credits

Per request

$0.004 · 0.4 credits

Each request is rounded up to 0.01 credits.

API

Call DeepSeek V4 Pro with your Railwail API key. Use this model ID in the request:

deepseek-v4-proAPI documentation Get an API key

curl https://railwail.com/api/v1/chat/completions \
  -H "Authorization: Bearer $RAILWAIL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-pro",
    "messages": [
      {
        "role": "user",
        "content": "Explain what a vector database is in two sentences."
      }
    ],
    "max_tokens": 1024
  }'

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["RAILWAIL_API_KEY"],
    base_url="https://railwail.com/api/v1",
)

completion = client.chat.completions.create(
    model="deepseek-v4-pro",
    messages=[
        {
            "role": "user",
            "content": "Explain what a vector database is in two sentences.",
        },
    ],
    max_tokens=1024,
)

print(completion.choices[0].message.content)

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.RAILWAIL_API_KEY,
  baseURL: "https://railwail.com/api/v1",
});

const completion = await client.chat.completions.create({
  model: "deepseek-v4-pro",
  messages: [
    {
      role: "user",
      content: "Explain what a vector database is in two sentences."
    }
  ],
  max_tokens: 1024
});

console.log(completion.choices[0].message.content);

// npm install railwail
import railwail from "railwail";

const rw = railwail(process.env.RAILWAIL_API_KEY);

const res = await rw.chat("deepseek-v4-pro", [
  { role: "user", content: "Explain what a vector database is in two sentences." },
], { max_tokens: 1024 });

console.log(res.choices[0].message.content);

Set your key as RAILWAIL_API_KEYCreate API key

Specifications

Model ID: deepseek-v4-pro
Developer: DeepSeek
Category: Text & chat
Input: Text
Output: Text
Context window: 1,048,576 tokens
Max. output: 384,000 tokens
Billing: By usage (tokens or GPU time)
Released: April 24, 2026
Lifecycle: Current version
Model size: 1.6T total / 49B active per token
License: Open weights under a permissive license that allows commercial use. Hosted API access via deepseek.com.
Catalog entry updated: September 23, 2026

Input parameters

Inputs and settings from the model's input schema. The example in the API section shows which of them the API accepts.

ParameterTypeDefaultAllowed values

promptrequired
User message
Type: Text
Default: –
Allowed values: up to 32,000 characters
top_p
Type: Number
Default: 1
Allowed values: 0 to 1
stream
Type: Yes/no
Default: false
Allowed values: –
max_tokens
Type: Integer
Default: 4096
Allowed values: 1 to 32,000
temperature
Type: Number
Default: 0.7
Allowed values: 0 to 2
system_prompt
Optional system instruction
Type: Text
Default: –
Allowed values: up to 8,000 characters

Use cases

What it is used for

Open-weights coding agents
Long-document analysis
On-premise enterprise deployments
Cost-efficient frontier reasoning workloads
Math and STEM tutoring backends
Research and academic use under permissive license
RAG over millions of tokens of context

Related guides

Best Large Language Models

All models in Text & chat

Frequently asked questions

What is DeepSeek V4 Pro?

DeepSeek V4 Pro is a model by DeepSeek in the Text & chat category. On Railwail you can call it with an API key through the Railwail API.

How much does DeepSeek V4 Pro cost on Railwail?

On Railwail, DeepSeek V4 Pro costs $1.584 per 1M input tokens and $4.752 per 1M output tokens. You are charged for what each request actually uses. Usage is paid from prepaid credits; 1 credit equals $0.01.

What is the context window of DeepSeek V4 Pro?

The context window of DeepSeek V4 Pro holds 1,048,576 tokens. One response can be up to 384,000 tokens long.

How fast is DeepSeek V4 Pro?

There are not enough measured runs of DeepSeek V4 Pro on Railwail yet to state a run time. It depends on the input, the settings and the load at the provider.

Is DeepSeek V4 Pro better than Claude Fable 5.1?

That depends on the task. DeepSeek V4 Pro (DeepSeek) and Claude Fable 5.1 (Anthropic) are both models in the Text & chat category. The comparison page shows their prices and specifications side by side.

Compare DeepSeek V4 Pro and Claude Fable 5.1

How do I use DeepSeek V4 Pro through the API?

Create a Railwail API key and send your request with the model ID deepseek-v4-pro. Code examples for curl, Python and JavaScript are in the API section of this page.

Comparable models

All in this category

Claude Fable 5.1Anthropic
Anthropic's model for the most demanding reasoning and long-horizon agentic work. 1M-token context window, up to 128K output tokens, adaptive thinking that is always on.
$12.00/1M in
658 % more expensive per unit
Compare DeepSeek V4 Pro vs. Claude Fable 5.1
Claude Opus 4.8Anthropic
The most capable model of Anthropic's Opus 4 series. State of the art on long-horizon agentic work, coding and knowledge tasks, with a 1M-token context window at standard pricing.
$6.00/1M in
279 % more expensive per unit
Compare DeepSeek V4 Pro vs. Claude Opus 4.8
Claude Opus 5.5Anthropic
Anthropic's current Opus model for long-running agentic coding and knowledge work. 1M-token context window, up to 128K output tokens, adaptive thinking that is always on.
$4.80/1M in
203 % more expensive per unit
Compare DeepSeek V4 Pro vs. Claude Opus 5.5

Use DeepSeek V4 Pro via the API

One API key for every model on Railwail. Usage is charged from prepaid credits, 1 credit = $0.01.

Get API key Browse all models