RSS

DeepSeek

DeepSeek V4 Pro

Benchmarks, price, context and versions

Also searched as

DeepSeek V4, DeepSeek V4 benchmark, deepseek v4 pro 0813, deepseek v4 pro price, deepseek v4 pro benchmark

Direct answer: DeepSeek V4 Pro is a model from DeepSeek. 91.7% on GPQA Diamond, per Epoch AI. On OpenRouter, US$ 0.96 per 1 million input tokens and US$ 1.91 per 1 million output tokens.

✓ Good for

  • technical reasoning on science topics (#30 of 198 on GPQA Diamond)
  • calculation and problems with an exact answer (#27 of 186 on OTIS Mock AIME)
  • fixing bugs in real code (#6 of 32 on SWE-bench Verified)
  • technical reasoning as the model appears through the API (#22 of 135 on GPQA Diamond (OpenRouter))
  • agents that follow rules and use tools (#16 of 131 on τ-bench Verified (airline))
  • very long documents or code (a window of 1 million tokens or more, according to OpenRouter)
  • running on your own infrastructure (open weights)

✕ Not the best for

  • Not enough data in the open sources this section uses to build this block.

This block follows a rule of this section, using source data: if the model is in the top 25% of the metric, the item goes under "Good for"; if it is in the bottom half, it goes under "Not the best for". The middle band appears in neither. Price and context use OpenRouter; scores use Epoch AI or OpenRouter Benchmarks, in the latest version we checked.

Benchmarks

GPQA Diamond

?what it is
91.7%

highest score on this metric 96.4%

30º / 198 · Epoch AI, latest version we checked

OTIS Mock AIME

?
98.6%

highest score on this metric 100%

27º / 186 · Epoch AI, latest version we checked

FrontierMath (tiers 1 to 3)

?
64.6%

highest score on this metric 95.1%

38º / 86 · Epoch AI, latest version we checked

SimpleQA Verified

?
52.9%

highest score on this metric 94%

26º / 87 · Epoch AI, latest version we checked

SWE-bench Verified

?
77.6%

highest score on this metric 83.5%

6º / 32 · Epoch AI, latest version we checked

GPQA Diamond (OpenRouter)

?
91.9%

highest score on this metric 96.5%

22º / 135 · OpenRouter Benchmarks, latest version we checked

τ-bench Verified (airline)

?
78.6%

highest score on this metric 82%

16º / 131 · OpenRouter Benchmarks, latest version we checked

Key specs

Origin
China
Family
DeepSeek V4 provisional data
Weights?
open
License
no data
Input US$/1M?
US$ 0.96
Output US$/1M
US$ 1.91
Context?
1,048,576 tokens
Parameters
no data
Input
text
OpenRouter usage?
0.66%
Downloads (30 days)?
no data
Use cases provisional data
Text Agents Code
Dates in sources
  • 2026, published, per Epoch AI
  • 2026, listed on OpenRouter

Version history

Dates come from the sources: OpenRouter listing or Hugging Face repo creation, which may differ from the official launch.

FAQ

How much does DeepSeek V4 Pro cost?

On OpenRouter, US$ 0.96 per 1 million input tokens and US$ 1.91 per 1 million output tokens.

Is DeepSeek V4 Pro open source?

Yes, it has open weights. License: no data.

What is DeepSeek V4 Pro's GPQA Diamond score?

91.7% on GPQA Diamond, per Epoch AI.

What is DeepSeek V4 Pro?

DeepSeek V4 Pro is a model from DeepSeek, with open weights, used for text, agents, code.

What is the context window of DeepSeek V4 Pro?

1,048,576 tokens, per OpenRouter.

Is DeepSeek V4 Pro good for coding?

It scores 77.6% on SWE-bench Verified, per Epoch AI.

Who made DeepSeek V4 Pro?

It was made by DeepSeek, a lab from China.

provisional data Questions ordered by what shows up in public Google search suggestions (no volume). Answers use source data only.

Compare with

Models from DeepSeek

See lab

Sources

  • Epoch AI, CC BY 4.0, latest version we checked.
  • OpenRouter, dados públicos da API; exports de benchmarks e rankings em CC BY 4.0, latest version we checked.
  • OpenRouter Benchmarks, CC BY 4.0, latest version we checked.

Dates come from the sources: OpenRouter listing or Hugging Face repo creation, which may differ from the official launch. Family and version were grouped by rule from the name.