How much does DeepSeek V4 Pro cost on ATP Token?

List rates for DeepSeek V4 Pro (API id deepseek-v4-pro): input $2.40 / output $4.80 per 1M tokens. Billing uses credits (1 USD = 100 credits), debited from the project balance per completed request. There is no monthly platform fee; see Pricing for the minimum top-up.

How do I call DeepSeek V4 Pro via API?

Create a project API key, enable the API id deepseek-v4-pro (not the display name) on that project’s allowlist, point an OpenAI-compatible SDK at https://api.atptoken.ai/v1, and set model to "deepseek-v4-pro". See the Quickstart on this page and the docs.

What are the context and capabilities of DeepSeek V4 Pro?

Vendor documentation lists a context window of about 1,000,000 tokens. Capability flags (thinking, tools, vision) follow the gateway model catalog — treat the catalog visible to your key as operational truth. Vendor: DeepSeek.

Who can use DeepSeek V4 Pro on ATP?

Only models on a project’s allowlist can be called with that project’s key (match the API id). Organization → workspace → project governance, per-request logs, and credit budgets are managed in the ATP Console.

DeepSeek V4 Pro

DeepSeek

DeepSeek V4 Pro is DeepSeek's flagship V4 MoE model — 1.6T total with about 49B active parameters per the vendor's release notes — aimed at heavy reasoning, coding, and agentic work at open-model economics. It is particularly strong at reasoning-dense coding, math and analysis tasks, 1M-context synthesis over large document or code sets, and long-horizon agents that alternate thinking and non-thinking turns. DeepSeek documents a 1M context with up to 384K output and thinking mode on by default, switchable off per request. On ATP Token you call the model id through a project key and pay list rates in credits under the same governance plane as closed providers.

ThinkingToolsVision

Get started free Full rate table

Modality

Text

Context

tokens

Input

$2.40

USD / 1M tokens

Cache read

$0.20

USD / 1M tokens

Released

2026-04

DeepSeek documents 1M context and up to 384K output for V4 Pro on its first-party service, with thinking mode default-on. Operational limits on ATP follow the gateway catalog for your project key.

List rates

USD list · 1 USD = 100 credits

Every completed request debits project credits at these list rates. Failed requests are not billed.

Input

$2.40

240 credits · USD / 1M tokens

Cache read

$0.20

20 credits · USD / 1M tokens

Output

$4.80

480 credits · USD / 1M tokens

Where it fits

Good fit

Reasoning-heavy coding and long-horizon agents on an open-model budget
1M-context synthesis and multi-file analysis
Math, science, and analytical tasks that benefit from default thinking mode
Mixed fleets routing open models beside closed APIs under one governance plane

Usually not ideal

Vendor-mandated OpenAI- or Anthropic-only compliance scopes
Bulk, latency-first traffic — V4 Flash is the throughput tier

On ATP

The same governance applies to every model on the gateway:

Project API key — spend and revocation bound to one project
Model allowlist — only enabled models may be called
Credits — 1 USD = 100 credits, debited per completed request
Per-request logs — model, key, project, credits for audit

Cost estimator

A list-price estimate for a month of traffic. Actual spend follows each completed request.

ESTIMATE

Monthly cost estimator

Input tokens / month10M

Output tokens / month2M

ESTIMATE

3,360

credits / month

$33.60

USD / month

List-price estimate (1 USD = 100 credits); actual cost comes from each request's response. Prompt-cache hits bill input at the lower cache-read rate, so real spend can be lower.

Quickstart

Point your existing SDK at the gateway and use a project key. Read the docs

quickstart.py · OpenAI SDK

from openai import OpenAI

client = OpenAI(
  base_url="https://api.atptoken.ai/v1",
  api_key="atp-•••"
)

r = client.chat.completions.create(
  model="deepseek-v4-pro",
  messages=[{"role":"user","content":"hi"}],
)▍

Deepseek V3 2deepseek-v3-2 · $0.57 / $1.71 DeepSeek V4 Flashdeepseek-v4-flash · $0.20 / $0.40 GPT-5.4 Progpt-5.4-pro · $30.00 / $180.00 Claude Fable 5claude-fable-5 · $10.00 / $50.00 GPT-5.6 Solgpt-5.6-sol · $5.00 / $30.00 GPT-5.5gpt-5.5 · $5.00 / $30.00

FAQ

How much does DeepSeek V4 Pro cost on ATP Token?: List rates for DeepSeek V4 Pro (API id deepseek-v4-pro): input $2.40 / output $4.80 per 1M tokens. Billing uses credits (1 USD = 100 credits), debited from the project balance per completed request. There is no monthly platform fee; see Pricing for the minimum top-up.
How do I call DeepSeek V4 Pro via API?: Create a project API key, enable the API id deepseek-v4-pro (not the display name) on that project’s allowlist, point an OpenAI-compatible SDK at https://api.atptoken.ai/v1, and set model to "deepseek-v4-pro". See the Quickstart on this page and the docs.
What are the context and capabilities of DeepSeek V4 Pro?: Vendor documentation lists a context window of about 1,000,000 tokens. Capability flags (thinking, tools, vision) follow the gateway model catalog — treat the catalog visible to your key as operational truth. Vendor: DeepSeek.
Who can use DeepSeek V4 Pro on ATP?: Only models on a project’s allowlist can be called with that project’s key (match the API id). Organization → workspace → project governance, per-request logs, and credit budgets are managed in the ATP Console.

Sources

Editorial claims on this page cite the sources below; rates come from the ATP list.

DeepSeek — V4 release notesas of 2026-08-04
DeepSeek API — models & pricingas of 2026-08-04
DeepSeek API — docs home / model idsas of 2026-08-04

Get started

Start with USD 5 — credits never expire.

Issue a project key, enable this model, and pay only for what you use.

Get started free Talk to enterprise