Ministral WebGPU: Run Mistral's new multimodal models 100% locally in your browser.

Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 15 comments

Today, Mistral released **Mistral 3**, a family of multimodal models, including three start-of-the-art dense models (3B, 8B, and 14B) and Mistral Large 3 (675B, 41B active). All Apache 2.0! 🤗 Surprisingly, the 3B is small enough to run 100% locally in your browser with WebGPU acceleration, powered by Transformers.js. Link to demo: [https://huggingface.co/spaces/mistralai/Ministral\_3B\_WebGPU](https://huggingface.co/spaces/mistralai/Ministral_3B_WebGPU)

Reply to Post

15 Comments

[-]

Cast_Iron_Skillet@reddit

Got roasted. https://preview.redd.it/fxsbtw4ugu4g1.png?width=702&format=png&auto=webp&s=e94926714d42e1d5d40e94a6e7dfee788ea26781

[-]

tirolerben@reddit

You have to look on the bright side of life: The dying process of the human body begins around the age of 25. From then on, more cells die than new healthy cells are produced. Basically, from the age of 25, one dies until one is dead. 38 is a respectable age, and you are still here, old man.

[-]

po_stulate@reddit

People reproduce at 10 in medieval times? Is that physically possible?

[-]

RosestemX3@reddit

If the females monthly has started, mhm And the men? I dont have experience in that category

[-]

KeinNiemand@reddit

Where the new 70B dense models? All just tiny dense models and huge MoEs

[-]

layer4down@reddit

From what I can tell, releases seem to be split between Mistral and Ministral at the moment, which is a bit confusing.

[-]

previse_je_sranje@reddit

Cool demo video, but it doesn't load on either my laptop or phone

[-]

HenkPoley@reddit

If you are using Microsoft Edge, you may need to disable "Enhanced security" for that website, to enable fast calculations to load.

[-]

MustardTofu_@reddit

The demo is being run using your local setup with WebGPU. Neither your phone and probably also not your laptop have the required hardware. It states so clearly in the description of this post.

[-]

SoggyYam9848@reddit

If you told me we'd have multimodal models being run locally back when ChatGPT4 just launched I would've asked for some of what you were smoking.

[-]

Pyros-SD-Models@reddit

3 years ago some guy in this sub told me I have no clue about machine learning and that I don’t work in AI because I said that we have realistic full HD video generation done before 2030 because according to them “reality is too complex and would need a completely different form of architecture” and people working in the field would understand this simple truth. This was especially funny because we already had a quite advanced video model prototype at this time and people in the field actually know that vision is way easier to scale than text since one modality is perception ally bounded and the other isn’t but what do I know.

[-]

Ministral WebGPU: Run Mistral's new multimodal models 100% locally in your browser.

Reply to Post

15 Comments

Cast_Iron_Skillet@reddit

tirolerben@reddit

po_stulate@reddit

RosestemX3@reddit

KeinNiemand@reddit

layer4down@reddit

previse_je_sranje@reddit

HenkPoley@reddit

MustardTofu_@reddit

SoggyYam9848@reddit

Pyros-SD-Models@reddit

Cast_Iron_Skillet@reddit

epSos-DE@reddit

Pristine-Woodpecker@reddit

egomarker@reddit