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

251–260 of 682 posts

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

#251

Post by fwipsy »

esafak wrote: post #211
fwipsy wrote: post #192

Earlier quoted context omitted.

Good points, I personally believe that if/when China takes the lead, they will immediately stop releasing model weights. It only makes sense as a strategy to counterbalance (current) American labs' monopoly on frontier models. Holding both those positions would be hypocritical all right, but are you sure it's the same people commenting/voting in both cases? I don't think there's a strong consensus on Hacker News. Eve…

Free models help move robots; a complement. I can foresee the models staying free.

Maybe you're right for smaller models, but for frontier models this is the tail wagging the dog. If China has exclusive frontier model capability in the future, that's an enormous commercial and geopolitical lever. They won't throw that away to sell a few more robots. Anything beyond their competitors' capabilities will remain closed.

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

#252

Post by aliasxneo »

ericmay wrote: post #171
avaer wrote: post #101

I lament the comments saying this in any way redeems Meta (the company). The researchers releasing this stuff have almost nothing to do with Meta other than being bankrolled by the slaughterhouse. You aren't the customer, you are the pawn in big tech's game of thrones. Your good will is a commodity to be traded, almost literally. It will be used against you the moment it's convenient. This is open weights because Met…

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…

I suspect it's just the generalized anti-West/anti-American sentiments extended into anything and everything. Anything that makes the US look anything close to good goes against their cause and therefore must be countered and talked down.

But you're not wrong about the bias here. You just don't see many comments talking about it because they get mass flagged/downvoted for obvious reasons.

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

#254

Post by aruggirello »

scrlk wrote: post #5

Will be interesting to see how Qwen3.8 27B compares against this once it releases this week. Seems like dense 30B is back in fashion? EDIT: An open weight version of Muse Spark 1.2 is going to be released as well: https://x.com/alexandr_wang/status/2086756152034066792 https://xcancel.com/alexandr_wang/status/2086756152034066792

> Seems like dense 30B is back in fashion?

Huh, well... no? Gemma A4B and Qwen A3B are quite popular in fact. I'm sure 3.8 35B A3B will outperform 3.6 27B by all metrics

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

#255

Post by dannyw »

HardCodedBias wrote: post #247
ignoramous wrote: post #48

Earlier quoted context omitted.

> Meta did not abandon opensource Open weights* I don't think outside of the Big 3 (Ant, OAI, GDM), given the strong competition from China, any other Lab has a chance at capturing the coding market if they aren't open weights (save for xAI whose latest Grok looks every bit good & will probably rely on Cursor for distribution instead of going open weights). There's literally no other selling point, as the capabilitie…

GDM -- Ok, I'll bite. Why are you including them?

Some labs go through bad patches, GDM is definitely in one right now and the recent departures are not reassuring, but I think it's too early and dismissive to count them out of the race so far. They just need one good frontier release for everyone to go "GDM is back!"

Claude models weren't really good or noteworthy until the 3 series anyway.

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

#256

Post by ekianjo »

IMTDb wrote: post #31
zoobab wrote: post #23

Earlier quoted context omitted.

"With dirt cheap models like deepseek-v4-flash that will run "forever" on $10, the answer for me is clearly: no." When it's free, you are the product.

Deepseek flash is open weight, this means we can download and run that model without any connection to deepseek, no data/tokens/usage data ever reaches them. They cannot make us their product.

Running Ds flash at acceptable speeds is challenging unless you have several thousands of dollars to invest

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

#258

Post by lostmsu »

xienze wrote: post #118
lostmsu wrote: post #92

Earlier quoted context omitted.

They say it is trained with quantization awareness, so it should only be 15GB or so. Qwen was only trained in FP8 with QAT. UPD, NVM, got misled by comments here. It is actually almost 60 GB so much larger

You're mixing up sizes of different quants. The 60GB is unquantized, and Qwen's unquantized size is around 54GB. Their sizes as like quantization levels are similar.

From my perspective it doesn't make sense to talk about the number of parameters. What matters is model size in bytes and its performance at that certain size.

Meta actually relesed official 4 bit quants in 17GB, but I haven't seen any indication that training was quant-aware, so the quants are not going to have same performance. 3.6 27B has official FP8 quant that AFAIR was trained with quantization awareness.

The best example is last year's gpt-oss which was released prequantized in mxfp4 so 20B parameter model was under 14GB and 120B was under 70GB right away.

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

#259

Post by Manfrednotfunny »

zbendefy wrote: post #194
Manfrednotfunny wrote: post #27

Earlier quoted context omitted.

I'm waiting for the speed/quality per dollar metric to go down a little bit further and then I will def run it at home. Its not just that you send a sentence to an API endpoint, you always send EVERYTHING to that agent as a context. You want to analyse your spending history? You now send everything to someone. Either no one cares but understands this implication on how easy it is to really capture you or no one reall…

Similiarly I wonder why we dont run our own email server despite the sensitive data there.

I did, it started to become too much work to run it well due to all the spam :| (even with the right signatures and configs, until you learn what a blacklisted ip is and that ips need some time of 'positive history' and what not.....)

But at least with your email, you had to trust only one company, as shitty as it is.

Separation of concerns was also easy.

Now with OpenRouter, you just might by accident, send your whole context to just everyone because OpenRouter just routes to different models and you might just switch around between some free model, the good one etc. And it is always the whole context.

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

#260

Post by exceptione »

avaer wrote: post #101

I lament the comments saying this in any way redeems Meta (the company). The researchers releasing this stuff have almost nothing to do with Meta other than being bankrolled by the slaughterhouse. You aren't the customer, you are the pawn in big tech's game of thrones. Your good will is a commodity to be traded, almost literally. It will be used against you the moment it's convenient. This is open weights because Met…

  > being bankrolled by the slaughterhouse.
Thanks, that was a very loud LOL.
Post reply on HN