AI Model Watch
Models / DeepSeek

DeepSeek-V4-Pro

Preview

Larger/most-capable V4 model (~1.6T total / ~49B active params per HuggingFace model card + authoritative third-party reports; MIT License; mixed FP4/FP8). Context length 1M, max output 384K tokens. Input price $0.435/M cache-miss, $0.003625/M cache-hit; output $0.87/M (USD). Supports three reasoning-effort modes (non-think / think high / think max), JSON output, tool calls; FIM completion non-thinking-mode only. Concurrency limit 500. Part of the 'DeepSeek V4 Preview' generation (released 2026-04-24), hence status=preview. Knowledge cutoff NOT officially published by DeepSeek -> left null. Prices re-verified to the cent against api-docs.deepseek.com/quick_start/pricing/ on 2026-08-09. VENDOR-DECLARED FORWARD PRICE EVENT, stated as footnote (2) on that page and still present 2026-08-09: 'We plan to raise the overall pricing for DeepSeek API services in the near future, with a significant increase expected. Please plan your usage accordingly. The specific pricing plan will be subject to official notice.' DeepSeek states NO date and NO figure, so no changelog entry and no rate change is recorded here - only the announcement itself. Sourced absence (2026-08-09): the page states no peak/off-peak differential; a 2x off-peak note read on this same page on 2026-08-01 is no longer present, and DeepSeek published no notice of its removal. Responses API NOT yet supported; footnote (1) commits to adding it in 'early August 2026' and the feature table still shows it unsupported as of 2026-08-09. The pricing page is served at the TRAILING-SLASH url (2026-08-11): the extensionless form returns a different document ("Your First API Call") with HTTP 200.

DeepSeek-V4-Pro by DeepSeek costs $0.435 per 1M input tokens and $0.87 per 1M output tokens ($0.544/1M blended), with a 1M (1.000.000-token) context window. It is in preview. Its weights are publicly downloadable, so it can be self-hosted on your own infrastructure.

Last verified: 13 Aug 2026 · sourced from official provider documentation

Provider
DeepSeek
Status
Preview
Input price
$0.435 / 1M tokens
Output price
$0.87 / 1M tokens
Cached input
$0.004 / 1M tokens
Blended price
$0.544 / 1M tokens
Context window
1.000.000 tokens (1M)
Max output
384.000 tokens
Modality
text
Open weights
Yes — self-hostable
Knowledge cutoff
—
Released
24 Apr 2026
API string
deepseek-v4-pro

Source: DeepSeek official documentation ↗

See also

Compare DeepSeek-V4-Pro head-to-head

DeepSeek-V4-Pro vs DeepSeek-V4-Flash→

$0.544 vs $0.175 blended /M · same family

FAQ

DeepSeek-V4-Pro — questions & answers

How much does DeepSeek-V4-Pro cost?

DeepSeek-V4-Pro is priced at $0.435 per 1M input tokens and $0.87 per 1M output tokens ($0.544/1M blended at a 3:1 input-to-output ratio), with cached input at $0.004 per 1M tokens.

What is the context window of DeepSeek-V4-Pro?

DeepSeek-V4-Pro has a 1.000.000-token context window (1M), with up to 384.000 output tokens per request.

Is DeepSeek-V4-Pro deprecated?

No — DeepSeek-V4-Pro is in preview and not currently scheduled for deprecation or retirement in our tracker.

Does DeepSeek-V4-Pro support image or vision input?

Our catalog lists DeepSeek-V4-Pro's modalities as text; image/vision input is not among them.

Can I self-host DeepSeek-V4-Pro?

Yes — DeepSeek-V4-Pro is an open-weight model: its weights are publicly downloadable, so you can run it on your own infrastructure (subject to its license) instead of only calling DeepSeek's API.

What is the API model string for DeepSeek-V4-Pro?

The API model identifier for DeepSeek-V4-Pro is "deepseek-v4-pro" when calling DeepSeek's API.