Researcher says six Chinese models show Nvidia Blackwell fingerprints
Original title: Former OpenAI researcher says he found Nvidia Blackwell fingerprints hidden inside six Chinese AI models
Will DePue says unused weights preserve Hopper or Blackwell core-count patterns and published verification code.
Former OpenAI researcher Will DePue says six Chinese models released since April 2026 carry Nvidia Blackwell initialization signatures, while earlier models such as DeepSeek-V3 and Kimi K2.5 match Hopper. He maps DeepSeek-V4-Flash and MiniMax-M3 to 148-core B200 or B300 chips, and DeepSeek-V4-Pro, Kimi K3, Kimi K3.1-Pro, and Ling-3.1-flash to 152-core GB200 or GB300 systems. He says the method does not show where the GPUs are or prove a US export-law violation.
Unconfirmed: The fingerprints are DePue's analysis; they show initialization hardware, not ownership, location, or a legal violation.
- He estimates five examined runs used about 12,000 Blackwell GPUs combined.
- Kimi K3 alone is estimated at 4,500 Blackwell processors over about 70 days.
- He says the method identified all 24 control models with known hardware.
- Qwen3.8 and models from Zhipu, Tencent, Xiaomi, and StepFun still show Hopper.
- Meituan's LongCat-2.0 points to Huawei hardware, according to the report.
- Vendors
- Nvidia
- Countries
- ChinaUnited States
From the original article
3 min read at the-decoder.com
Former OpenAI researcher Will DePue says he has produced the first technical evidence that DeepSeek, Moonshot AI, MiniMax, and Ant Group are using Nvidia's latest Blackwell GPUs despite US export controls. His forensic method reads characteristic chip signatures directly from published model weights. He estimates the hardware behind the five training runs cost more than half a billion dollars.
By Maximilian Schreiner · the-decoder.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.