nando gaarenstroom

IT Support Engineer | Local AI Enthusiast
Exploring open-source LLMs, AI agents, building, testing, and learning.

[nando gaarenstroom @Nando_G_AI]

Holy!!! Look at those benchmark scores🔥🔥 GPT-6 Astra is going to be amazing!

[Chubby♨️ @kimmonismus]

Full benchmarks. Absolutely insane.

ARGI-AGI 3, from 7.8% to 98.6%

[nando gaarenstroom @Nando_G_AI]

I hope this can also help older cards as I still have an old ggz 1060 6gb which is catching dust. Could be good if I can use that for the mtp or dflash part of the models.

[NVIDIA RTX Spark @NVIDIARTXSpark]

Your devices are stronger together. 🖥️🤝🖥️

Just announced at IFA, NVIDIA PAIR automatically links systems across your local network and sends inference requests wherever there’s available capacity, helping agents run more efficiently.

[nando gaarenstroom @Nando_G_AI]

@iamEvanYT @MiaAI_lab @IFM_AI Really? In some benchmarks it does perform similarly but in terminal bench 2.1 for example it got a 14 point lead. That’s why I am excited about the model + the 512k context. I am unable to try the model for myself at the moment but will do it tomorrow.

[nando gaarenstroom @Nando_G_AI]

@MiaAI_lab I have filled it in and I hope I can atleast help in some way!

[nando gaarenstroom @Nando_G_AI]

@TeksEdge @Acemagic_ I was looking at a dgx spark but this one from amd has really captured my attention. The 50% increase in ram is something that will really help with bigger models.

[nando gaarenstroom @Nando_G_AI]

@ThePrimeagen Damn not Fortnite AI😭😭

[nando gaarenstroom @Nando_G_AI]

We didn’t yet get a qwen3.8 35b moe but we do have another now called K2-Horizon-MoVA-36B-A4B. I am interested to see how good it compares to qwen!

[Institute of Foundation Models @IFM_AI]

Introducing K2 Horizon: a connected fleet of six foundation models ranging from 0.9 billion to 375 billion parameters.

- Frontier performance: Across coding and agentic tasks, K2 Horizon delivers top-tier performance in every size class—with the 0.9B, 3.7B and 7B models setting new state of the art at their respective scales.
- Radical openness: K2 Horizon represents the largest fully open-source model launch in AI history. The fully open code, training data and recipes are a significant step forward in transparency.

Launch page: https://t.co/gg0k803SbL
Tech blog: https://t.co/g35L5xMGdS
Hugging Face: https://t.co/3Lb28JhyG9

[Institute of Foundation Models @IFM_AI]

Introducing K2 Horizon: a connected fleet of six foundation models ranging from 0.9 billion to 375 billion parameters.

- Frontier performance: Across coding and agentic tasks, K2 Horizon delivers top-tier performance in every size class—with the 0.9B, 3.7B and 7B models setting new state of the art at their respective scales.
- Radical openness: K2 Horizon represents the largest fully open-source model launch in AI history. The fully open code, training data and recipes are a significant step forward in transparency.

Launch page: https://t.co/gg0k803SbL
Tech blog: https://t.co/g35L5xMGdS
Hugging Face: https://t.co/3Lb28JhyG9

[nando gaarenstroom @Nando_G_AI]

It has finally been done! Huggingface has been officially bought by nvidia for 13 billion dollars! Lets hope nvidia will not let us down.

[Jensen Huang @JensenHuang]

Exciting day for NVIDIA and @huggingface.

Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. They allow every developer, startup, university, industry and country to build with, customize and benefit from AI.

Thank you @ClementDelangue for coming to me.

NVIDIA is going to be a great home for Hugging Face, its community and the future of open models. 🤗

https://t.co/q8Om2Xc5ye

[nando gaarenstroom @Nando_G_AI]

@superalesha I have actually seen you post and then came across his and was confused. I thought I was reading the same thing and got Déjà vu. Anyways glad to actually find the actual owner!

[nando gaarenstroom @Nando_G_AI]

@chacon Yep sadly European models are just really bad. It would’ve been good to have some competition but they are just easily forgotten. Might also be because of the laws they have to follow in the EU.

[nando gaarenstroom @Nando_G_AI]

@RemiFabreRobot @ClementDelangue @antoinepirrone I mean from what I have seen it works really wel and looks fun to use!

[nando gaarenstroom @Nando_G_AI]

@JuanCereigido It looks like a very promising project! I hope it really does well as this can really help older generations!

[nando gaarenstroom @Nando_G_AI]

@maria_rcks No tag with AI generated so it must be real!

[nando gaarenstroom @Nando_G_AI]

@PaulGugAI Yep it doesn't make the benchmark look good. But it also makes Anthropic look bad as the models are all over the place.

[nando gaarenstroom @Nando_G_AI]

@Thom_Wolf This is truly what we were missing in the AI ecosystem a funny duck on a swing! With every post I see on this fellow I get more and more the urge to also buy one.

[nando gaarenstroom @Nando_G_AI]

@0xSero Thank you! Sadly 300b is not something I can run (yet).

[nando gaarenstroom @Nando_G_AI]

@Chris_Wozniczek @0xSero Thanks! I had a feeling the base Qwen3.6 was better than the rest but when looking at the reported benchmarks I was unsure.