Skip to content
NIST·

🧪NIST Evaluates DeepSeek V4 Pro: Trails US Frontier Models by Eight Months but Wins on Cost

NIST Evaluates DeepSeek V4 Pro: Trails US Frontier Models b…

TL;DR

The Center for AI Standards and Innovation at NIST published its May 2026 evaluation of DeepSeek V4 Pro.

The Center for AI Standards and Innovation at NIST published its May 2026 evaluation of DeepSeek V4 Pro. Capabilities trail leading US closed models by roughly eight months and roughly match GPT-5, while DeepSeek V4 proved more cost-efficient than GPT-5.4 mini on five of seven benchmarks.

Key Points

1

Capabilities trail US frontier models by ~8 months

2

Open-weight performance similar to GPT-5

3

More cost-efficient than GPT-5.4 mini on 5 of 7 benchmarks

Why It Matters

First official US government benchmark of a major Chinese open model, with direct implications for export controls and open-source policy.

deepseeknistbenchmarkschina-ai

Frequently Asked Questions

Why does this matter?

First official US government benchmark of a major Chinese open model, with direct implications for export controls and open-source policy.

What happened?

The Center for AI Standards and Innovation at NIST published its May 2026 evaluation of DeepSeek V4 Pro.

Comments

Subscribe to join the conversation...

Be the first to comment

Enjoyed this article?

Get it daily. 7am. Free. Reads in 5 minutes.

Join 3,461 builders reading daily.

Also get