# Hey New Mistral model incoming

**URL:** <https://forums.developer.nvidia.com/t/hey-new-mistral-model-incoming/385255>\
**Category:** DGX Spark / GB10\
**Tags:** mistral-large, deepseek\
**Created:** [October 7, 2026, 12:41am UTC](https://forums.developer.nvidia.com/t/hey-new-mistral-model-incoming/385255 "2026-10-07T00:41:51Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![WilliamD](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/williamd/32/452493_2.png) [@WilliamD](https://forums.developer.nvidia.com/u/WilliamD)\
**Post date:** [October 7, 2026, 12:41am UTC](https://forums.developer.nvidia.com/t/hey-new-mistral-model-incoming/385255/1 "2026-10-07T00:41:51Z")

</div>

Hey,

When Mistral announced deploying and serving GLM-5.3, even though it is an excellent model, it felt like the end… Well guess what, a new model dropped and pretend to be a direct contender to deepseek-v4.1. The weights are available at the end of this month.

> **[Introducing Mistral Large 4 | Mistral](https://mistral.ai/news/mistral-large-4/)**
>
> The most powerful AI platform for enterprises. Customize, fine-tune, and deploy AI assistants, autonomous agents, and multimodal AI with open models.

Well, I am quite surprised they still release a model trying to reach the level of Chinese models (even if a bit late). I was pleased with the previous Mistral coding model (Devstral) but felt like in 2026 they couldn’t keep up.

Hope this one isn’t all smoke. I can’t run it locally because of its 1T parameters, but I would still give it a try via OpenRouter eventually. I’m mostly interested in cyber applications.

Fingers crossed.  
William