My 16 year old daughter just got advice from her Pediatrician on how to get into Medical School - go to a small college where the Profs know you and you can convince them to inflate your grades, there is too much competition in big schools and no room to influence your grades outside your work So basically, try to cheat a little. Thanks Doc.

Well, being female, gives her an advantage for Med School....AND if she's a minority female, that really helps mover her closer to the top of the list....

It's only one semester, and a semester of the least important material of their university time.

Err...where did you go to school where 1st semester/year wasn't important stuff?

I was in all the weed out classes....chemistry, calculus 1...etc....

Hell I didn't get any "easy" classes like sociology or one English reading classes (sci fi books) till like I was a Junior or maybe senior.....

"In order to salve the mental health of freshman student..."

JFC...are you fscking KIDDING ME?

Is this just the next natural progression from everyone getting a participationi trophy in grade schools?

Why is it today's kids are so mentally deficient in ways that all other generations till the past decade haven't been troubled with??

Look, not everyone is cut out for college ....and those early classes, are weed out classes like others have said...at least they used to be.

Stress is a real part of REAL LIFE and this is a good time to learn these lessons.

Learn them early if they aren't college material....and let them get out before racking up massive debt in a degree that won't pay shit in the end.

I read the other day where incoming freshman these days, I think in the 5%-8% range had to take remedial math?!?! Our 1-12 schools aren't preparing them.....

If we keep lowering the standards....we're not helping anyone, we're hurting all them in the long run.

Let's re-think educations....some kids SHOULD be left behind for higher education, but that doesnt' mean we don't teach them some real skills where they can earn a living and contribute to society.

But for God's sake, let's quit coddling these little fuckers...they're going soft in ways that never even occurred to previous generations....

One of the requirements to become a US Citizen is you have to show some proficiency in American English.

Other countries have an official language, we need to make it official here.

We want people coming into the US to assimilate and the greatest tools for that is common language.

You rarely go to other countries and hear commercials in different languages ......certainly the signage for streets, etc is not in multiple languages...etc.

There is nothing bigoted about expecting someone trying to live in your country to speak the common language ....

The USA doesn't have an official language.

We need to fix that...other countries have official languages.....why not us?

Here in the US, audio recording without consent is prohibited by wiretap laws, but not video recording... sigh.

That's not quite true....

That's ONLY true in states that mandate both parties consent to and know they're being recorded.

Many, if not most states...allow for one party consent, so usually the person recording knows its recording and therefore legal.

If you record in public without consent, the only thing you should be allowed to do with that video is maintain it in private and use it for security or investigative purposes for a fixed time window. Any other use should be prohibited, and yes that this includes use by the media.

News reporting?

And anyone can be a news reporter potentially.....

This is the USA.

Speak fucking English!!!

1) 30B, dense. Optimized to fit Q4 quantized in a 24GB card with a speculative decoding model as well.

2) Because reporters don't know what weights are and assume you don't know either.

3) Way better than Gemma 4 on text tasks, slightly better on multimodal. Numbers below are all: Benchmark: Muse Glimmer score Gemma 4 31B score difference

Artificial Analysis Intelligence Index: Muse Glimmer: 35 30 +5
MCP Atlas (Public): 75.5 54.2 +21.3
DeepSearch QA: 74.6 61.7 +12.9
SWE-Bench Pro: 51.2 36.9 +14.3
SWE-Bench Verified: 76.0 66.6 +9.4
OSWorld-Verified: 65.9 58.5 +7.4
GAIA2: 43.3 36.4 +6.9
WildClawBench: 47.6 37.6 +10.0
TerminalBench 2.1: 51.7 43.4 +8.3
Tau3-Banking: 23.5 15.1 +8.4
MMMU Pro: 74.0 73.0 +1.0
Charxiv Reasoning: 78.8 77.7 +1.1
OmniDocBench v1.5: 75.8 72.5 +3.3
ScreenSpot Pro: 75.4 75.9 -0.4

4) 128k tokens

5) No, sadly.

What you want is a DGX Spark.

It costs $4000.

And yes, there are fundamental advantages to cloud services, such as large-scale batching, little idle downtime, high speed, and hardware optimized to the specific models / serving needs. That said, one can weigh that off against sovereign control over your server...

Define "tolerable".

The model in question - Muse Glimmer - with DFlash/speculative decoding - will probably get you ~60 to 124 tok/s on a 3090 (a quite dated GPU). For a 5090, it's said to clock in at 233,4 tok/s. On CPU you're looking at maybe 3-5 tok/s.

If you call that "tolerable", I guess you're more patient than me? And as mentioned, you're not just wasting time, but also wasting a lot of power too - CPU is a very power-inefficient way to run ML models.

If you insist on CPU, this isn't the right kind of model anyway. You want to take advantage of the fact that you probably have lots of (comparably cheap) RAM, and compensate for the fact that you have (comparably) terrible memory bandwidth, and for that, you want a MoE with a high total parameters but a low active parameters. Not a dense model like this.

Then I was basically aiming at Mac Minis

That's very much a special case which you didn't mention in your post that I responded to, but still the answer is "meh". You couldn't run it at all on a 16GB Mac Mini, and I think you'd struggle to run it at all on a 24GB (because you have to share the ram with the OS, the inference server, etc). For the base Mini you might get 10-12 tok/s, and for the M2 Pro / M4 Pro, maybe 25-35 tok/s. Still pretty far from a GPU, though.

This model is designed for >= 24GB GPUs.

MoEs don't save RAM (for a given quality), they increase it. You have to store all of the parameters in memory, not just the active params. But inference only uses a subset of the total params for each token, so it reduces the memory bandwidth requirements and improves token generation rate. But this comes at the cost of a higher total param count for a given quality.

Meta has always been releasing open models (Llama was famously the first powerful open model). The change has actually been in the opposite direction, with Muse Spark *not* being open.

Zuck descrbed his motivation way back when, about how they got burned with Facebook on app stores, in that Apple and Google could basically bully them however they wanted, on whatever extractive terms they wanted, and there was nothing Meta could do about it. He's now paranoid about "others controlling the platform", and wanted to make sure that doesn't happen with AI, that they have their own AI base to work with.

I mean, if you want your model to be terribly slow and terribly power inefficient, sure, go ahead and run it on CPU.

This really isn't a good model for CPU. For CPU, you want a MoE with a large number of total params but a tiny number of active params. Something like DeepSeek V4 Flash 0731 if you have at least 128GB of RAM - you might get 2-3 tok/s or so on that. The goal is to minimize the memory bandwidth requirements per token, at the cost of a greater total RAM footprint.

For GPU, you're highly VRAM limited but not bandwidth limited, so your best option is generally a dense model (non-MoE) with speculative decoding to make up for the performance limitations.