DeepSeek-V4-Pro-Preview

The high-capability DeepSeek V4 preview: a 1.6T-total, 49B-active MoE with a 1M context window.

LegacyLanguageReasoningCodingAgentOpen weights

MODEL NUMBER019

Specifications

Only published specifications are shown.

Total parameters

1.6T

Active parameters

49B

Architecture

MoE

Context

Input modalities

Text

Open weights

Yes

License

DeepSeek Model License

Status

Legacy

The high-capability DeepSeek V4 preview: a 1.6T-total, 49B-active MoE with a 1M context window.

It introduced V4’s sparse-attention, million-token architecture and released the weights alongside the preview.

PreviousDeepSeek-V3.2671B · 128K ctx

CurrentDeepSeek-V4-Pro-Preview1.6T · 1M ctx

New capabilities Adjustable reasoning

DSA

Adjustable reasoningAgentCoding

2026-04-24Preview

DeepSeek released V4-Pro Preview with open weights.

Select a card to view the related model.