Few hours ago, Intern Team: https://chat.intern-ai.org.cn came out with their latest most powerful AI model. It got my attention because of the huge gap on benchmarks with the top frontier AI models worldwide claiming benchs of 55% against 6-13% only in Biology for the giants evaluated in "trillions".

Decided to test it immediately. All I needed was Converting the Safetensors to GGUF and Quantizing to the sweet spot Quantization that make a LLM model affordable hardware-quality wise.

---

Recently has been a Dilemma if certain intelligence especially in Chemistry and Biology should be available to masses. Is all year 2026 that many experts are warning about slowing down the AI pace development because reaching unknown/uncontrollable levels of Intelligence but also because what may happen if this Intelligence goes in the wrong hands.

---

Intern_S2 is Qwen 397B based, which many knows about it´s capabilities. It has evolved, same way like GPT became GPT2-3-4 etc other actors are evolving not certainly to underestimate. The real danger is twice threating:

• 1 The use for malicious purposes.
• 2 The opensource mopping the floor with frontier LOL (disrupting current tech giants revenues, and an entire future economical ecosystem).

---

The big Question:

Whos to decide what here.

There are bad actors everywhere, that's why "distributing" intelligence responsibly is a must, but is not that the tech giants are trustable much more. Then Again:

Who put anyone in charge of such an important matters, who gave the right to decide about it, who decides what, in an unregulated novel tech? Already we start seeing the first cracks even in the most controlled companies Like OpenAi/Anthropic and even Meta with "AI agents going Rogue". This may be a advertisement stunt but again who really knows. Simply loosing credibility with irresponsible actions. Then pushing to apply AI to Electricity Power and Water Grid? Scary... The facts show that the only companies acting responsibly so far (in the west) are Mistral and (I hate to admit this one but is the truth) Google.

Getting back to Intern-s2 AI model, it is already available for use with HugstonOne Enterprise Edition

and the LLm model will be available (the q4_k_m) for download for 1-7 days, in Hugston.com. It sizes ~240 GB and it can run in CPU/GPU.

All it is still being tested, with great excitement.