openopen-weightsReleased 2026-04

DeepSeek V4 Pro

By DeepSeek · Near-flagship reasoning at low cost.

Input / 1M

$0.43

Output / 1M

$0.87

Context

1.0M tokens

Max output

384K tokens

DeepSeek V4 Pro pricing

DeepSeek V4 Pro costs $0.43 per million input tokens and $0.87 per million output tokens at standard API rates, with cached input at $0.015/M. That makes it cheaper than 20 of 25 other tracked models on input.

Estimate your monthly cost →

DeepSeek V4 Pro benchmarks

SWE-bench Verified80.6%

GPQA Diamond90.1%

MMLU-Pro87.5%

LiveCodeBench93.5%

Ranks #4 on SWE-bench Verified among tracked models.

Specs

Maker: DeepSeek
Category: open
Released: 2026-04
Modalities: text
Open weights: Yes (MIT)
Context window: 1.0M tokens
Max output: 384K tokens

Official API docs ↗

Compare DeepSeek V4 Pro with

Mistral Large 3 (2512)

Mistral AI · $0.50/$1.50 per 1M

Qwen3.5-397B-A17B

Alibaba · open-weights

Kimi K2.6

Moonshot AI · $0.67/$3.50 per 1M

GLM-5.2

Z.ai · $1.40/$4.40 per 1M

← All AI models Cost calculator Which model should I use?

Data verified 2026-06-20. Prices are standard USD API rates per 1M tokens.