DeepSeek V4 Pro

V4 Pro is the large model in DeepSeek's V4 family and the pick when you want an open-weight option that holds up in agent workflows. It is a mixture-of-experts design with 1.6 trillion parameters in total and 49 billion active for each token. A preview came out in April 2026 and the general release in August 2026, with the main improvements in how it behaves as an agent in production.

The weights are public. On Kadraris you call it with the same key as closed models, which makes head-to-head tests simple.

What you can use it for

• Agent pipelines that were built and tuned on V4 Pro
• Reading whole codebases or document collections in one go
• Very long generated pieces, such as complete reports or specs
• Testing an open-weight model against closed ones on your own data

Settings and limits

• Context window: 1,000,000 tokens; output up to 384,000 tokens.
• DeepSeek defines three thinking effort levels for it: low, high and max.
• Output is priced noticeably higher than input, so long generations make up most of the cost.
• Cap max_tokens when you only need a short answer.
• Call it through /v1/chat/completions with the OpenAI SDK.

Example request

from openai import OpenAI

client = OpenAI(
    base_url="https://kadraris.com/v1",
    api_key="sk-kd-...",
)
resp = client.chat.completions.create(
    model="deepseek-v4-pro",
    messages=[{"role": "user",
               "content": "Write a design doc for a retry "
                          "queue with dead-letter handling."}],
)
print(resp.choices[0].message.content)

Replace the topic with your own system and add constraints such as language or framework.

Questions

How much does V4 Pro cost?

$0.54 / $1.22 per 1M tokens on Kadraris. Watch output length, since that is where most of the spend goes. A request that fails is not charged.

Is V4 Pro going away?

DeepSeek first said it would route V4 Pro traffic to its newer Flash model, then changed course after user feedback and kept V4 Pro live. A V4.1 Pro has been mentioned with no public date, so keep a migration plan ready.

Should I start with Flash instead?

Often yes. DeepSeek V4 Flash comes from the same family and is several times cheaper on both input and output. Keep V4 Pro for tasks where it wins on your data or where existing prompts depend on it. See also GLM 5.3 vs DeepSeek V4 Pro.

DeepSeek V4 Pro in your code

from openai import OpenAI

client = OpenAI(base_url="https://kadraris.com/v1", api_key="sk-kd-...")
resp = client.chat.completions.create(
    model="deepseek-v4-pro",
    messages=[{"role": "user", "content": "Suggest three opening shots for a short film about a cafe."}],
)
print(resp.choices[0].message.content)

curl https://kadraris.com/v1/chat/completions \
  -H "Authorization: Bearer $KADRARIS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "deepseek-v4-pro", "messages": [{"role": "user", "content": "Suggest three opening shots for a short film about a cafe."}]}'

import OpenAI from "openai";

const client = new OpenAI({ baseURL: "https://kadraris.com/v1", apiKey: process.env.KADRARIS_API_KEY });
const resp = await client.chat.completions.create({
  model: "deepseek-v4-pro",
  messages: [{ role: "user", content: "Suggest three opening shots for a short film about a cafe." }],
});
console.log(resp.choices[0].message.content);