---
title: "DeepSeek V4 Pro API Pricing & Context | RZ Automate"
description: "DeepSeek V4 Pro costs $1.32 per 1M input and $3.96 per 1M output tokens at peak, half that off-peak. 1M context, MIT open weights."
url: https://rzautomate.com/models/deepseek-v4-pro
---

[All AI models](https://rzautomate.com/models)

[DeepSeek](https://rzautomate.com/models?provider=DeepSeek#models)

Flagship model

# DeepSeek V4 Pro API pricing and specs

DeepSeek V4 Pro costs $1.32 per million input tokens and $3.96 per million output tokens at peak hours. Off-peak, the price drops by half. DeepSeek V4 Pro reads up to 1 million tokens, writes up to 384K, and ships as open weights under the MIT license.

[Try the cost calculator](#calculator)
[Compare models →](#compare)

* Tool calling
* Reasoning
* Structured output
* Open weights

Input

$1.32 / 1M

Output

$3.96 / 1M

Context

1M tokens

Max output

384K tokens

Pricing

## DeepSeek V4 Pro API pricing

Output costs 3× input. Run big jobs off-peak and the bill halves.

Peak hours run 1 to 4 am and 6 to 10 am UTC, Monday to Friday. Every other hour and all weekend is off-peak.

Prices checked November 14, 2026 on DeepSeek's pricing page.

__DeepSeek V4 Pro API pricing per 1M tokens in US dollars__
| Price type   | Peak per 1M | Off-peak per 1M |
| ------------ | ----------- | --------------- |
| Input        | $1.32       | $0.66           |
| Cached input | $0.044      | $0.022          |
| Output       | $3.96       | $1.98           |

Calculator

## DeepSeek V4 Pro cost calculator

Pick a job or enter your own numbers. We use peak rates so the estimate stays safe.

Start from a job

1,000 support emails100 invoices500 leads

Rate period

NowOff-peak

Requests per monthInput tokens per requestOutput tokens per requestShare of input from cache (%)

Uses peak rates. Off-peak halves the result. Ignores tax and cache savings unless you set them.

Per month

$4.22

Cost per request

$0.0042

Per year

$50.69

Same job on other models

* [DeepSeek V4.1 Flash$1.08](https://rzautomate.com/models/deepseek-v4-1-flash)
* [Muse Spark$4.20](https://rzautomate.com/models/muse-spark)
* [Kimi K3$12.00](https://rzautomate.com/models/kimi-k3)

* [AI customer support agents](https://rzautomate.com/services/ai-customer-support-agents)
* [AI document processing](https://rzautomate.com/services/ai-document-processing)
* [AI lead generation & CRM automation](https://rzautomate.com/services/ai-lead-generation)

Want this costed on your real data? [Book a free 30-min call](https://rzautomate.com/book?service=agent-development&source=models&model=deepseek-v4-pro)

Specs

## DeepSeek V4 Pro context window and specs

Context window

1M tokens

Max output

384K tokens

__DeepSeek V4 Pro specifications__
| Spec              | Details                                 |
| ----------------- | --------------------------------------- |
| Input types       | Text                                    |
| Output types      | Text                                    |
| Tool calling      | Yes                                     |
| Structured output | Yes                                     |
| Reasoning         | Thinking and non-thinking modes         |
| Knowledge cutoff  | Not published                           |
| Release date      | April 24, 2026 (update August 13, 2026) |
| Size              | 1.6T parameters, 49B active             |
| Open weights      | Yes                                     |
| License           | MIT                                     |
| API model ID      | deepseek-v4-pro                         |

Availability

## Where to access DeepSeek V4 Pro

The weights are open, so many hosts serve it. Third-party hosts set their own prices and skip the peak schedule.

__Platforms that offer DeepSeek V4 Pro__
| Platform     | Model ID                      | Input $/1M     | Output $/1M    |
| ------------ | ----------------------------- | -------------- | -------------- |
| DeepSeek API | deepseek-v4-pro               | $1.32          | $3.96          |
| OpenRouter   | deepseek/deepseek-v4-pro-0813 | Varies by host | Varies by host |
| Fireworks AI | —                             | $1.74          | $3.48          |

Free access

## Is DeepSeek V4 Pro free?

The DeepSeek V4 Pro API is paid. There is no free API tier. The weights are free to download under MIT, so you can run it on your own hardware.

* No free API tier
* Free weights under MIT
* Self-hosting needs serious GPUs

Alternatives

## DeepSeek V4 Pro vs alternatives

DeepSeek V4 Pro is the cheapest open-weight flagship here. For simple jobs V4.1 Flash does the work for about a quarter of the cost.

__DeepSeek V4 Pro compared with alternatives__
| Model                                              | Input $/1M | Output $/1M | Context | Why pick it                      |
| -------------------------------------------------- | ---------- | ----------- | ------- | -------------------------------- |
| DeepSeek V4 Pro                                    | $1.32      | $3.96       | 1M      | Open weights, strong reasoning   |
| [DeepSeek V4.1 Flash](https://rzautomate.com/models/deepseek-v4-1-flash) | $0.30      | $1.20       | 1M      | Same family, far cheaper         |
| [Muse Spark](https://rzautomate.com/models/muse-spark)                   | $1.25      | $4.25       | 1M      | Reads images, video and PDFs     |
| [Qwen3.8-Max](https://rzautomate.com/models/qwen-3-8-max)                | $2.00      | $6.00       | 1M      | Multimodal flagship from Alibaba |
| [Kimi K3](https://rzautomate.com/models/kimi-k3)                         | $3.00      | $15.00      | 1M      | Top open-weight agent scores     |

[See all AI models →](https://rzautomate.com/models)

Fit

## What DeepSeek V4 Pro is good for

### A good fit for

* Multi-step agents on a tight budget
* Long documents up to 1M tokens
* Teams that may self-host later

### Not the best pick for

* Images, scans or PDFs as images
* Strict data residency outside China
* High-volume simple tasks

Our take

We use DeepSeek V4 Pro when an agent needs real reasoning and the budget is tight. We run batch jobs off-peak.

* [AI agent development →](https://rzautomate.com/services/ai-agent-development)

History

## DeepSeek V4 versions

The V4 Pro ID now serves the August 13 checkpoint. Old V3 aliases deepseek-chat and deepseek-reasoner are retired.

1. [DeepSeek V4 Pro (0813)](https://rzautomate.com/models/deepseek-v4-pro)Latest  
2026-08-13  
$1.32 in / $3.96 out per 1M
2. [DeepSeek V4.1 Flash](https://rzautomate.com/models/deepseek-v4-1-flash)Latest  
2026-09-10  
$0.30 in / $1.20 out per 1M
3. DeepSeek V4 ProPrevious  
2026-04-24  
$0.435 in / $0.87 out per 1M
4. DeepSeek V4 FlashDeprecated  
2026-04-24  
$0.44 in / $1.32 out per 1M
5. DeepSeek V3.2Deprecated  
2025-12-01  
$0.28 in / $0.42 out per 1M

Questions

## DeepSeek V4 Pro FAQ

Prices checked November 14, 2026.

### How much does DeepSeek V4 Pro cost?

DeepSeek V4 Pro costs $1.32 per million input tokens and $3.96 per million output tokens during peak hours. Off-peak rates are half, so $0.66 and $1.98\. Cached input costs $0.044 at peak. Answering 1,000 support emails a month costs about $4.22 at peak rates.

### Is DeepSeek V4 Pro free?

The DeepSeek V4 Pro API is not free and has no free tier. You pay per token, with lower prices off-peak. The model weights are free under the MIT license, so you can download DeepSeek V4 Pro and run it yourself if you have the GPUs for a 1.6 trillion parameter model.

### What is the DeepSeek V4 Pro context window?

DeepSeek V4 Pro has a 1 million token context window and can write up to 384K tokens in one reply. That fits whole contracts, codebases or long email threads in a single call. DeepSeek V4 Pro charges the same rate across the full window, with no long-context surcharge.

### When was DeepSeek V4 Pro released?

DeepSeek V4 Pro was first released on April 24, 2026\. DeepSeek shipped an updated checkpoint, DeepSeek V4 Pro 0813, on August 13, 2026, and the API now serves that version. Peak and off-peak pricing for DeepSeek V4 Pro started on August 16, 2026.

### What is the DeepSeek V4 Pro API model ID?

The DeepSeek V4 Pro API model ID is deepseek-v4-pro on the DeepSeek API. On OpenRouter it is listed as deepseek/deepseek-v4-pro-0813\. DeepSeek once planned to route this ID to V4.1 Flash, but cancelled that, so deepseek-v4-pro still serves the Pro model.

### DeepSeek V4 Pro vs DeepSeek V4.1 Flash, which is cheaper?

DeepSeek V4.1 Flash is cheaper. It costs $0.30 input and $1.20 output per million tokens at peak, against $1.32 and $3.96 for DeepSeek V4 Pro. For 1,000 support emails that is about $1.08 a month on Flash versus $4.22 on DeepSeek V4 Pro.

## Want DeepSeek V4 Pro working inside your tools?

We build AI agents and automations that let a business grow without adding headcount. On the call we tell you which model fits your work and what it costs at your volume.

* Email  
[choaib@rzautomate.com](mailto:choaib@rzautomate.com)
* Reply time  
Within two hours
* 160automations running in production
* 32businesses served
* 2,000 hrsof manual work taken off clients every month

Or book directly at [cal.com/rzautomate/30min](https://cal.com/rzautomate/30min)

Want the full picture? See [AI agent development](https://rzautomate.com/services/ai-agent-development).

Keep comparing

## Compare more models

* [DeepSeek V4.1 FlashSame family at a quarter of the price](https://rzautomate.com/models/deepseek-v4-1-flash)
* [Kimi K3Open-weight flagship from Moonshot](https://rzautomate.com/models/kimi-k3)
* [Qwen3.8-MaxAlibaba's multimodal flagship](https://rzautomate.com/models/qwen-3-8-max)

1. [Home](https://rzautomate.com/)
2. [AI Models](https://rzautomate.com/models)
3. [DeepSeek](https://rzautomate.com/models?provider=DeepSeek#models)
4. DeepSeek V4 Pro
