Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows

research.meta.ai

Post reply on HN
Print view View original

281–290 of 682 posts

Re: Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows

#281

Post by NorwegianDude »

claytongulick wrote: post #229
staplor wrote: post #205

Earlier quoted context omitted.

What do you mean iffy? The major AI labs are gross profitable when selling access to inference. In addition, the best models have trillions of parameters and are most efficiently served on large, expensive clusters and served to many concurrent users.

> The major AI labs are gross profitable when selling access to inference. Do you have a good source for this?

Well, I can run some models that are better than some of the weaker and cheaper Anthropic models locally, like Haiku 4.5, and solve tasks that would cost ~4500$ every day in tokens, so yeah, they are definitely extremely profitable on inference.

Re: Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows

#282

Post by cmiles8 »

staplor wrote: post #205
cmiles8 wrote: post #97

With the business model for API based LLMs looking iffy at best it seems like we’re heading back to the “server under your desk” era of IT again.

What do you mean iffy? The major AI labs are gross profitable when selling access to inference. In addition, the best models have trillions of parameters and are most efficiently served on large, expensive clusters and served to many concurrent users.

That’s like saying an apartment building is “profitable” because the rent covers utilities while ignoring the real cost which is the mortgage on the capital cost of the building.

It’s funky math and a good way to quickly go bankrupt.

Re: Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows

#284

Post by mark_l_watson »

Meta is rocking AI. As of last week I have been using their excellent muse coding harness with their model Muse Spark 1.2.

Starting this morning I am running their new local 30B model muse-glimmer on my old MacMini 32G using Ollama (remember to increase the context size!) and pi coding harness. I am getting good results with muse-glimmer running locally, with the caveat that everything runs slowly (e.g., give it a task and then go walk outside or do Qi Gong exercises for a while).

Re: Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows

#285

Post by rufasterisco »

exe34 wrote: post #220
skohan wrote: post #134

Earlier quoted context omitted.

I've been coding using the LLM server in my living room for the past few weeks, and I haven't had this much fun with tech for ages

Can I ask, do you feel the pain of the level of abstraction? I haven't tried local in a few months, but last time I tried, I felt like I was directing a coding exercise - whereas with a frontier model, it feels more like directing a product building. "I need this feature", vs "write code to do this in this file".

Some people like it better when they direct the solution because they walk away with a better understanding of it.

This has emotional/psychological aspects (it feels less like LLMs are replacing you), as well as practical ones (overall complexity is bounded by what the dev brain can understand/grasp).

A dev work becomes more and more about reliability, signing off safe software with a litmus test: “I will be on to handle this code failure as if I had written it”.

All the above points towards keeping tight control over some level of abstractions and delegating others.

Re: Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows

#286

Post by applfanboysbgon »

ericmay wrote: post #235
tarr11 wrote: post #222

Earlier quoted context omitted.

Had to look this up. https://en.wiktionary.org/wiki/Goomba_fallacy

Not applied accurately with respect to my comment, but it is a funny one and also new to me in the naming.

> It’s rather amusing to me to read comments like this, and then simultaneously whenever a Chinese company or team releases open-weight models or whatever there is a giant round of applause, America is so behind, and there’s nothing but positive things to say about the intelligent, creative, and well-intentioned Chinese engineers

It is absolutely applied accurately. You're commenting on the alleged hypocrisy of people simultaneously criticising American open-source while praising Chinese open-source, and then attributing your perception of hypocrisy to the website as a whole. The reality is the behaviour you've observed comes from completely different individuals, not some kind of HN hivemind. Your comment is such a typical case that it could go in the wiktionary page as the example excerpt teaching people what the goomba fallacy is.

Re: Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows

#287

Post by ericmay »

throwup238 wrote: post #276
ericmay wrote: post #202

Earlier quoted context omitted.

> I don't think there's a strong consensus on Hacker News. Even something like the time of day an article is posted might get different engagement depending on who is active in which time zones. Based on my own experience and reading, I do think there's a general consensus on this site but I could certainly be wrong about that. I'm less concerned about hypocrisy per se, it's more that the arguments that are used, eve…

> It's common, if not inevitable, for people who feel strongly about $topic to conclude that the system (or the community, or the mods, etc.) are biased against their side. One is far more likely to notice whatever data points that one dislikes because they go against one's view and overweight those relative to others. This is probably the single most reliable phenomenon on this site. Keep in mind that the people wit…

I'm certainly pro-America and anti-communist/fascist as a bias, but I don't really care about whether AI tools are open-weight or not. I just use the products that best fit my needs. I just think the arguments put forth regarding China and open-weight models and strategy are not very good. It just so happens that China is the only real competitor in the AI space and so they are who get talked about the most in comparison to the United States.

This comment isn't applicable to me, and if you believed that it applied, you'd have to add it to the OP as well since they feel strongly about Meta[1], they notice data points about Meta's behavior, and they overweight their bias against Meta[1] relative to others. Same with China "leading" and open-source/open-weight models and any time someone says China's strategy is better.

You can repeat this for any online argument or any topic.

It's not that Dang is wrong, however. It's that posting it in response to my comment(s) alone is hypocritical and pointless. Whereas Dang who is more responsible for the entire community is right to speak about it more generally. The message matters but so does the messenger, in this case.

[1] I don't use any Meta products (I don't even click on links), think social media should probably be outright banned, and Meta very likely should have been sued into the ground for the effects that their platform seems to have not just on children and young adults but also on our political system.

Re: Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows

#288

Post by ninjin »

InfiniteLoup wrote: post #9
tosh wrote: post #2

good to see new open weights releases from meta

The least they could do, after ruthlessly bombarding my employer's servers with requests, ignoring the robots.txt, scraping everything, and incurring significant Google Maps costs for us in the process.

2a03:2880: by any chance?:

https://news.ycombinator.com/item?id=48137854

Have asked them to stop numerous times and they just keep hitting for about eight months now.

Re: Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows

#289

Post by spaceywilly »

mark_l_watson wrote: post #284

Meta is rocking AI. As of last week I have been using their excellent muse coding harness with their model Muse Spark 1.2. Starting this morning I am running their new local 30B model muse-glimmer on my old MacMini 32G using Ollama (remember to increase the context size!) and pi coding harness. I am getting good results with muse-glimmer running locally, with the caveat that everything runs slowly (e.g., give it a ta…

Newb question but I’m curious what would help it to run faster? Would it need more vRAM or just system memory?

Re: Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows

#290

Post by andelink »

dominotw wrote: post #212
ericmay wrote: post #171

Earlier quoted context omitted.

It’s rather amusing to me to read comments like this, and then simultaneously whenever a Chinese company or team releases open-weight models or whatever there is a giant round of applause, America is so behind, and there’s nothing but positive things to say about the intelligent, creative, and well-intentioned Chinese engineers (which is true, America certainly doesn’t have a monopoly on great people). Don’t you know…

There is a big astroturfing going on social media platforms by the chinese. Did you notice 'day in a life of unmarried 30 yr old lady in china' videos flooding usa social media. Regular ppl in the west now hold mildly positive views of the ccp and how 'advanced' china is than usa. Then there are europeans who now are looking for china to give them the technology handout now that relationship with usa has soured.

I only see those type of videos on 小红书
Post reply on HN