It's a bit scary! Does everyone know what kind of strength the scythe you're trading against in the stock market has?

Let's start with some news.

The domestic AI large model DeepSeek unexpectedly exposed version 3, which has quite strong performance, ranking second in the Aider multilingual testing leaderboard with a score of 48.4, surpassing Claude3.5 sonnet and only behind OpenAI o1's score of 61.

In the LiveBench evaluation, DeepSeek v3 is the strongest open-source large language model, ranking second among non-inference models.

So, who developed such an impressive large model? Many might think it was an internet giant.

Wrong! This model was developed by Huanfang Quantization, yes, the one involved in quantitative trading in the country.

It is widely rumored that Huanfang is the company with the most high-performance GPU cards in the country.

Additionally, according to media reports: Xiaomi is building a GPU cluster with tens of thousands of cards, increasing its investment in AI large models; previously, Xiaomi already had 6,500 GPUs.

The latest key figure who joined Xiaomi to lead the large model team is Luo Fuli, a key developer of DeepSeek v2.

Now everyone knows what kind of opponent they are trading with, saying it's a nuclear-powered scythe is not an exaggeration at all.

Honestly, I am both happy about the progress of domestic large models and worried for the investors; remember: the sell orders from 1 to 5 opposite you are likely international leading AI large models, so everyone should be cautious.