# LM STUDIO very week quality with great models

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Flm-studio-very-week-quality-with-great-models%2F139782%3E%5C
**Category:** Models\
**Created:** [February 9, 2025, 12:23am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Flm-studio-very-week-quality-with-great-models%2F139782 "2025-02-09T00:23:27Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![languedeboeuf](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/languedeboeuf/32/40727_2.png) [@languedeboeuf](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Flanguedeboeuf)%5C
**Post date:** [February 9, 2025, 12:23am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Flm-studio-very-week-quality-with-great-models%2F139782%2F1 "2025-02-09T00:23:27Z")

</div>

I downloaded LM STUDIO and the mistral 7B 0.3 model, which is a recognized model. However, the answers I get in LM are really bad…  
Why such a difference with the answers on the mistral website?

---

<div class="post-metadata">

**Author:** ![nyuuzyou](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/nyuuzyou/32/40724_2.png) [@nyuuzyou](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fnyuuzyou)%5C
**Post date:** [February 9, 2025, 1:51am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Flm-studio-very-week-quality-with-great-models%2F139782%2F2 "2025-02-09T01:51:44Z")

</div>

Le Chat by Mistral uses a different model, the [Mistral Large](https://huggingface.co/mistralai/Mistral-Large-Instruct-2411) (if I’m not mistaken). 7B 0.3 is the legacy version.

But this is definitely not a model that will run on a home computer,

> Note: Running Mistral-Large-Instruct-2411 on GPU requires over 300 GB of GPU RAM.

---

<div class="post-metadata">

**Author:** ![languedeboeuf](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/languedeboeuf/32/40727_2.png) [@languedeboeuf](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Flanguedeboeuf)%5C
**Post date:** [February 9, 2025, 10:20am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Flm-studio-very-week-quality-with-great-models%2F139782%2F3 "2025-02-09T10:20:52Z")

</div>

Thanks. it seems to me that the answers are incomparably worse than even the first online versions. I have the same observation after testing other models. I wonder if some of the local LLM users get good results? All LLMs combined.

---

<div class="post-metadata">

**Author:** ![languedeboeuf](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/languedeboeuf/32/40727_2.png) [@languedeboeuf](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Flanguedeboeuf)%5C
**Post date:** [February 9, 2025, 11:31am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Flm-studio-very-week-quality-with-great-models%2F139782%2F4 "2025-02-09T11:31:03Z")

</div>

I realise the answers are not so bad when I type them in english. Previously I wrote them in french
