Is DeepSeek V4 Coming? No, It Is Already Live

If you are still searching “Is DeepSeek V4 coming?”, the short answer has changed:

• DeepSeek-V4 is already live in preview.
• DeepSeek announced the V4 release in April 2026.
• The current public API model names are deepseek-v4-pro and deepseek-v4-flash.

This article keeps the original question because people still search for it, but the factual answer is now clear from DeepSeek's own public pages.

The quick answer

• Is DeepSeek V4 coming? No. It has already been released in preview.
• Where is it available? Web, official app, API, and open-weight releases referenced by DeepSeek.
• What should developers use? deepseek-v4-pro or deepseek-v4-flash.
• Where should you verify updates? DeepSeek's release page, changelog, pricing page, and main product site.

What is officially confirmed

DeepSeek's public release page states that DeepSeek-V4 Preview is officially live and describes two models in the family:

1. DeepSeek-V4-Pro

   • Positioned as the stronger flagship option in the V4 family.
   • Listed with 1.6T total parameters and 49B active parameters on the release page.

2. DeepSeek-V4-Flash

   • Positioned as the faster, cheaper option.
   • Listed with 284B total parameters and 13B active parameters on the release page.

The same release note also says the API is available now and that both models support 1M context.

Release timing: the safe way to state it

The safest wording is: DeepSeek-V4 was publicly released in April 2026.

Why phrase it that way instead of forcing one exact timestamp?

• The DeepSeek changelog groups the V4 API update under 2026-04-24.
• The release announcement and product surfaces were published around the same release window.

For a guide site like this one, “released in April 2026” is accurate without pretending we have a more precise timestamp than the public pages consistently expose.

The model names you should actually use

For new API integrations, the public docs say to use:

• deepseek-v4-pro
• deepseek-v4-flash

DeepSeek also documents two older names:

• deepseek-chat
• deepseek-reasoner

Those older names still exist for compatibility, but the docs mark them for deprecation on July 24, 2026. DeepSeek says they currently map to the non-thinking and thinking modes of deepseek-v4-flash.

What the official API docs say

The current public API docs describe a setup that is intentionally simple:

• OpenAI-format base URL: https://api.deepseek.com
• Anthropic-format base URL: https://api.deepseek.com/anthropic
• Supported current V4 models: deepseek-v4-pro, deepseek-v4-flash

The public quick start also shows that you can use a standard OpenAI SDK client and point it to DeepSeek's base URL, instead of building a custom client stack from scratch.

Context length, output, and supported features

According to DeepSeek's public pricing page, both V4 models are listed with:

• 1M context length
• 384K maximum output
• Thinking and non-thinking modes
• JSON output
• Tool calls
• Chat prefix completion

The same table also lists FIM completion as beta support for non-thinking mode only.

Pricing that is publicly documented

DeepSeek publishes V4 pricing directly in the API docs.

At the time this article was updated, the page lists:

• DeepSeek-V4-Flash

  • $0.14 per 1M input tokens on cache miss
  • $0.0028 per 1M input tokens on cache hit
  • $0.28 per 1M output tokens

• DeepSeek-V4-Pro

  • $0.435 per 1M input tokens on cache miss
  • $0.003625 per 1M input tokens on cache hit
  • $0.87 per 1M output tokens

DeepSeek also notes a promotional pricing adjustment on the pricing page, so this is a category where readers should always re-check the official source before making cost decisions.

What we are not claiming

A lot of low-quality AI landing pages get rejected because they start filling gaps with marketing guesses. This site should not do that.

So here is what we are not claiming without a direct DeepSeek source:

• We are not claiming a broader multimodal feature set for V4 unless DeepSeek documents it clearly.
• We are not inventing enterprise governance features that are not spelled out on the public pages we reviewed.
• We are not treating third-party demos or screenshots as official DeepSeek infrastructure.

That matters more than it sounds. Accuracy is part of content quality.

Where to verify things yourself

If you want the primary sources, start here:

1. The DeepSeek V4 release page:
   • https://api-docs.deepseek.com/news/news260424
2. The DeepSeek changelog:
   • https://api-docs.deepseek.com/updates
3. The DeepSeek pricing page:
   • https://api-docs.deepseek.com/quick_start/pricing
4. The DeepSeek product site:
   • https://www.deepseek.com/en/

Those are the pages this guide should follow when the model lineup or pricing changes.

Bottom line

DeepSeek V4 is not a rumor anymore. It is already live in preview.

If you are integrating today, the practical takeaway is straightforward:

• use deepseek-v4-pro or deepseek-v4-flash
• keep https://api.deepseek.com as the base URL
• treat deepseek-chat and deepseek-reasoner as compatibility aliases, not the long-term target
• verify pricing and changelog details on DeepSeek's own pages before you ship

That is the cleanest factual answer to the original question.