DeepSeek V4-Pro — Spezifikationen
| Entwickler | DeepSeek |
|---|---|
| Typ | LLM (MoE) |
| Modality | Text → Text |
| Parameter | 1,6 Bio. insgesamt / ~49 Mrd. aktiv (MoE) |
| Kontextfenster | 1 Mio. |
| Maximale Ausgabe | 384 K |
| Lizenz | MIT (offen) |
| Offene Gewichte | Ja |
| Veröffentlicht | 2026-04 |
| Eingabepreis | 0,435 $ pro 1 Mio. |
| Ausgabepreis | 0,87 $ pro 1 Mio. |
| API-Anbieter | DeepSeek, OpenRouter |
Lokal ausführen
| VRAM (4-Bit) | ~800 GB |
|---|---|
| Mindest-GPU | Multi-GPU-Server (z. B. 8× H100 mit 80 GB) |
What is DeepSeek V4-Pro?
DeepSeek V4-Pro is the company’s open flagship: a 1.6-trillion-parameter mixture-of-experts
activating around 49B parameters per token, with a 1M-token context window and MIT-licensed
weights. The API speaks both OpenAI and Anthropic request formats, which makes it unusually
easy to drop into an existing codebase — often a base-URL and key change rather than a
rewrite.
Pricing is $0.435 in / $0.87 out per million tokens, which blends to roughly $0.52 — around
a twentieth of frontier pricing for a model that holds its own on general reasoning. That
ratio, not the parameter count, is why V4-Pro matters. The dual-format API compatibility
compounds it: the switching cost that normally protects incumbent providers largely
disappears, so V4-Pro is a genuine option for teams that would otherwise never evaluate a
Chinese lab’s model. Self-hosting is a data-centre exercise — roughly 800 GB of VRAM at
4-bit, so eight H100 80GBs or more — meaning the open licence here buys auditability and
provider portability rather than a realistic on-premises deployment for most organisations.
DeepSeek V4-Pro pricing: API cost per 1M tokens
| Eingabe (pro 1 Mio. Token) | $0.435 |
|---|---|
| Ausgabe (pro 1 Mio. Token) | $0.870 |
| Output/input ratio | 2× |
| Blended (4:1 in:out) | $0.522 per 1M tokens |
What DeepSeek V4-Pro costs per month
Real monthly spend at a 4:1 input-to-output mix — the ratio a typical chat or RAG workload actually produces.
| Workload | Tokens/Monat | Cost / month |
|---|---|---|
| Nebenprojekt | 1 Mio. Eingabe / 0,25 Mio. Ausgabe | $0.65 |
| Kleines Team | 20 Mio. Eingabe / 5 Mio. Ausgabe | $13 |
| Produktion | 200 Mio. Eingabe / 50 Mio. Ausgabe | $131 |
Run your own numbers in the KI-API-Kostenrechner.
Cheaper alternatives to DeepSeek V4-Pro
| Modell | Gewichteter Preis pro Million US-Dollar | You save |
|---|---|---|
| DeepSeek V4-Flash offenere | $0.168 | 68 % günstiger |
Self-host or pay the API?
DeepSeek V4-Pro is open-weight, so you can run it yourself. It needs ~800 GB of VRAM at 4-bit (Multi-GPU server (e.g. 8× H100 80GB)). Self-hosting only beats the API once your volume is high enough to keep that hardware busy — the Selbsthosting-vs.-API-Rechner works out the break-even point for your token volume.
Häufig gestellte Fragen
How much does DeepSeek V4-Pro cost per 1M tokens?
DeepSeek V4-Pro costs $0.435 per 1M input tokens and $0.870 per 1M output tokens. At a typical 4:1 input-to-output mix that blends to about $0.522 per 1M tokens.
How much does DeepSeek V4-Pro cost per month?
A small-team workload of 20M input and 5M output tokens a month costs about $13 on DeepSeek V4-Pro. A side project (1M in / 0.25M out) costs roughly $0.65.
What is a cheaper alternative to DeepSeek V4-Pro?
DeepSeek V4-Flash is the strongest cheaper option in our database at $0.168 per 1M blended — about 68% less than DeepSeek V4-Pro. It is also open-weight, so self-hosting is an option.
Can I run DeepSeek V4-Pro locally?
Yes. DeepSeek V4-Pro is open-weight and needs about ~800 GB of VRAM at 4-bit quantisation (Multi-GPU server (e.g. 8× H100 80GB)).
Why does DeepSeek V4-Pro charge more for output than input?
Output tokens are generated one at a time and cannot be batched the way a prompt can, so they cost the provider more to serve. DeepSeek V4-Pro charges 2× more for output, which is why prompt-heavy workloads are far cheaper to run than generation-heavy ones.
See every DeepSeek model priced side by side: DeepSeek API pricing.
Prices are the published list rates for the model's primary API and are reviewed as providers change them. Volume, batch and cached-input discounts are not included. Compare every model side by side in the Datenbank für KI-Modelle oder das LLM-Leaderboard.
