TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Mistral AI launches Mixtral-Next

204 pointsby varunvummadiover 1 year ago

15 comments

lolinderover 1 year ago
Mistral&#x27;s process for releasing new models is <i>extremely</i> low-information. After getting very confused by this link I tried looking for a link that has <i>any</i> better information, and there just isn&#x27;t one.<p>I thought Mixtral&#x27;s release was weird when they just pasted a magnet link [0] into Twitter with no information, but at least people could download and analyze it so we got some reasonable third-party commentary in between that and the official announcement. With this one there&#x27;s nothing at all to go on besides the name and the black box.<p>[0] <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=38570537">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=38570537</a>
评论 #39411773 未加载
评论 #39414786 未加载
评论 #39411790 未加载
mattpavelleover 1 year ago
For those unfamiliar with the LMSys interface:<p>Click&#x2F;tap on &quot;Direct Chat&quot; in the top tab navigation and you can select &quot;mistral-next&quot; as model.
评论 #39409402 未加载
评论 #39410118 未加载
twobitshifterover 1 year ago
AIExplained on youtube has guessed that Gemini 1.5 pro is taking Mistral’s accurate long content retrieval and Google just scaled it as much as they could. The Gemini 1.5 pro paper has a citation back to the last mistral paper in 2024.
评论 #39411424 未加载
mrfakenameover 1 year ago
Note that it&#x27;s actually &quot;Mistral Next&quot; not &quot;Mixtral Next&quot; - so it isn&#x27;t necessarily a MoE. For example, an early version of Mistral Medium (Miqu) was not a MoE but instead a Llama 70B model. I wonder how many parameters this one has
评论 #39411516 未加载
bloopernovaover 1 year ago
Slightly related question: what&#x27;s a good coding LLM to run on a 4070 12GB card?<p>Also, do coding LLMs use treesitter to &quot;understand&quot; code?
评论 #39410613 未加载
评论 #39411358 未加载
评论 #39410424 未加载
评论 #39412247 未加载
评论 #39411938 未加载
aunetxover 1 year ago
It&#x27;s quite funny to use! It is better when speaking French than chat gpt3.5 on my opinion
评论 #39411350 未加载
justanotherjoeover 1 year ago
wow, this might be the best LLM that i&#x27;ve used in terms of phrasing and presenting the answers.
评论 #39417287 未加载
AnujNayyarover 1 year ago
No indication that this a MoE (Mistral not Mixtral).<p>Very exciting nevertheless, here’s hoping the bless the OS community once again!
评论 #39417320 未加载
ismailmajover 1 year ago
Could it be Mistral Large? This beats GPT-4 on my personal test.
评论 #39410313 未加载
评论 #39409970 未加载
tmikaeldover 1 year ago
This was linked randomly on Mistrals Discord chat, nothing &quot;official&quot; yet.<p>It&#x27;s a preview of their newest prototype model.<p>To use it, click &quot;Direct Chat&quot; tab and choose &quot;Mistral next&quot;
ccwilson10over 1 year ago
I used this but, upon asking which model it is, it replied as being a &quot;fine-tuned version of GPT 3.5&quot;. Any clue why? In a second chat it replied &quot;You&#x27;re chatting with one of the fine-tuned versions of the OpenAssistant model!&quot;.
评论 #39422834 未加载
vitorgrsover 1 year ago
From my tests, it did better than Gemini Ultra on a few reason&#x2F;logic questions.
apapapaover 1 year ago
The Together.AI logo at the bottom is very hard to read... (Dark gray on black)
redder23over 1 year ago
You can literally type &quot;woke shit&quot; in and you get woke shit out. I am so impressed.
xeckrover 1 year ago
As someone who has only been using GPT-4 since its release, I am pleasantly surprised by how far open LLMs have come.