Our Smallest, Most Efficient Model: Muse Ember

Posted by dvaplima 6 hours ago

Counter229Comment96OpenOriginal

Comments

Comment by doodlesdev 16 minutes ago

Interesting to notice that the best benchmarks of the model are not on Meta's own agent harness, but instead on mini-swe-agent:

   > 1. All models run at max reasoning effort. Artificial Analysis results are from Intelligence Index v5, as published November 9, 2026; cost per task uses each model's list pricing. Internal results are from Meta's evaluation suite. The Personal Superintelligence Index (PSI) is an internal composite of eight evaluations: Agents' Last Exam, AutomationBench, FocusBench, JobBench, OfficeQA Pro, GDP.pdf, GDP.xlsx and VBT. Terminal-Bench 4.0 uses each model's best published harness: mini-swe-agent (Muse Ember), Claude Code (Claude Haiku 5.5), and Codex (GPT-6 Luna). OSWorld 2.1 uses partial-credit reward; PerceptionBench is scored with tools.

Comment by simonw 5 hours ago

Six effort levels (minimal, low, medium, high, xhigh, max), so here are six pelicans (actually five, "high" didn't return a complete SVG):

https://tools.simonwillison.net/markdown-svg-renderer?url=ht...

There's usually a reasonably strong correlation between thinking level and visual quality of the pelican. This is the first time I've seen that correlation appear to break - more reasoning tokens doesn't seem to produce an obviously better SVG.

Comment by meric_ 5 hours ago

Wow, minimal seems to be the best here interestingly. Only 10 reasoning tokens so basically none?

Comment by ezst 4 hours ago

Wouldn't that be a sign that those models were trained to optimise that test case specifically, as in, by exhibiting equal or higher performance while using less power?

Comment by ourcat 2 hours ago

Indeed. And that they were all trained to give it a red scarf, which was not asked for.

Comment by 9dev 1 hour ago

At this point, even if the labs wouldn’t actively pelican-max their models, isn’t the training corpus full of pelicans already? Funny to think how one models choice of adding a red scarf in one of Simon’s earlier experiments might have led to an archetype of pelicans riding bicycles…

Comment by dvaplima 2 hours ago

Its funny how to the naked eye, changing the effort seems like only a difference on styles

Comment by imimayj1337 5 hours ago

It does seem pretty insane to me that only on max setting (and after waiting 27 seconds) did you actually manage to receive an SVG where it appears that the pelican's legs are straddling the bike correctly.

Edited to make clear what it looks like to me and what I was looking for.

Comment by swatcoder 5 hours ago

They're not. It's an illusion from the composition.

If you look at the code for each of these (and probably most across models), you can see that the drawing elements are organized by concept. Presumably, this helps accuracy by keeping attention on one concept at a time, but you end up with a stack of conceptualized objects and lose the ability to interleave things.

No doubt these could all do better with a more thorough and specific prompt that coached them away from that technique, but otherwise it's a very subtle detail that a "thinking" process needs to stumble upon independently.

Comment by flir 5 hours ago

You could use a second prompt to z-order the objects (I tend to break down tasks and pipeline them, rather than trying to stuff everything in a one-shot prompt. Seems to work better for me, could be a holdover from the early days).

Comment by latexr 5 hours ago

They’re not. Note how both pedals are on the same side of the frame. That’s wrong!

Comment by eterm 5 hours ago

Did minimal just hand it off to something else? It seems completely out of line with the rest.

Comment by permalac 5 hours ago

Is there a problem if I like the minimal more than the others?

Comment by MadameMinty 5 hours ago

I like medium best, but minimal is not bad at all.

Comment by mring33621 4 hours ago

Yeah minimal has style

Comment by AlecSchueler 4 hours ago

Any idea why there's so much convergence on the fish in a basket idea?

Comment by swatcoder 4 hours ago

Same reason they're all facing right with the same composition, they all have three motion lines, they all have a yellow sun and 2-3 blobby clouds, they're all on two-lane pavement with a white dashed lane separator.

And the same reason Claude has all its notorious Claudisms in prose.

The shorter and vaguer a prompt you give to an LLM, the more it fills in further details using associations they've developed through training. These are ley lines that the models gravitate towards. Diffusion models too.

It's why lazy AI use for creative work always comes across as lazy in the wild and always will.

When you use AI for "creative" purposes yourself, it feels like magic to see a pelican riding a bicycle at all, and you're impressed by all those cool details it added. But then 10, or 100, or 10M people throw more or less the same trivial prompts at it and we start to recognize all those "ley line" associations fill in the gaps the same way again and again and again.

Comment by viraptor 3 hours ago

To be fair, people have a genuine preference for images with things facing to the right. And lots of drawings will do 3 motion lines. You'd likely draw a dashed rather than solid separator too, or it would look like a random line. Lots of the AI habits are there because of choices we make.

Comment by mvkel 4 hours ago

This is interesting. There's a balance between:

- simple prompt for complex output (bad)

- complex prompt for simple output (bad)

- simple prompt for simple output (good)

- complex prompt for complex output (.... is this good or bad?)

Comment by Buttons840 2 hours ago

There's only one average in a set.

Comment by phyrex 5 hours ago

I like that from xhigh on up its wearing a helmet.

Comment by santoshalper 2 hours ago

Wearing a helmet is smart, so that checks out.

Comment by sjacob 5 hours ago

Incredible. How do you get these out so fast?

Comment by simonw 2 hours ago

I was at my dog's sniffing class for this one, but I had Codex on my Mac (controlled from Codex on my iPhone) run the test prompt with my `llm-meta-ai` plugin, which has the ability to refresh to fetch new models that Meta AI have released: https://github.com/simonw/llm-meta-ai#usage

Comment by victorhooi 25 minutes ago

Sorry I have to ask....um....sniffing class? What is that?

Btw, loved your work with Django!

Comment by dofm 25 minutes ago

> dog's sniffing class

There's a blog post in this for sure. And a classic joke.

Comment by mgdev 4 hours ago

Fidelity.

Comment by jolt42 5 hours ago

Pelican is the new teapot

Comment by djmips 5 hours ago

birds at minimal and return at max

Comment by small_model 4 hours ago

Not sure this is a good way to benchmark LLM's, no offence, but a single SVG is hardly a measure of a models capability.

Comment by AmazingTurtle 5 hours ago

[dead]

Comment by Retro_Dev 3 hours ago

When they said the model was the "smallest" of their models yet, I was really hoping it would be open weights. I can't find any weights, how sad :(

Comment by loeg 3 hours ago

Spark Open was just October. Give it time. https://research.meta.ai/blog/muse-spark-open

Comment by kristianp 1 hour ago

Which was 816B parameters.

Comment by loeg 14 minutes ago

We don't know how many parameters the closed Meta models have, so what are you comparing this number against?

Comment by anon373839 1 hour ago

It was also nerfed to the point of no longer being in a comparable intelligence class to the original.

Comment by minimaxir 4 hours ago

> It costs $0.10 per million input tokens, $0.004 per million cached input tokens and $0.50 per million output tokens,

Notable cache cost of 4% input, compared to Luna/Haiku's 10% of input.

Comment by system2 2 hours ago

No batch. Pass.

Comment by kbrannigan 4 hours ago

This is starting to feel just like the early PC days. A bunch of Computer models coming out, all shapes and color, but eventually it settled around mouse keyboard, monitor

most people, use the office suite: Word, Spreadsheet , presentations, database.

I bet we're going to see the exact same thing with LLMs at specialized models instead of the everything models.

One for lawyers, coders, Accountants, Designers, etc…

Comment by GuB-42 52 minutes ago

Human lawyers, coders, accountants and designers are a lot more than their jobs.

They are people with common knowledge, a lawyer has to understand carpentry to understand how to deal with how a client's roof fell off, a coder has to understand transportation for writing a travel app, an accountant has to understand medicine to do accounting for a pharmacy, a designer has to understand cooking to design a toaster. Not expert knowledge, but enough to understand that toasters are hot and need electricity.

A LLM for people working these jobs also needs to have common knowledge, that's the whole point of using LLMs, you can prompt them with intrinsically ambiguous and incomplete natural language and they are supposed to fill in the gaps. And the more common knowledge they have, the better they are at it. So why settle for less? If it is an optimization problem, we already have the solution, it is the "mixture of experts", which is better at optimizing then having an "lawyer LLM", "coder LLM", etc... because it is better adapted to the internal structure of the neural network behind it.

Comment by grim_io 4 hours ago

I don't think so.

Cross discipline skills are more valuable in the real world, outside of hyper specialization seen in the California tech scene. Even there, it's coming to an end.

Comment by drillsteps5 2 hours ago

Exactly what I was thinking. If this model will be used for coding, does it need to be fed the entire (mostly pirated) collection of works from Anna's Archive? Probably not. Maybe some, primarily computer-science oriented.

The main corpus would be a collection of thoroughly reviewed, clean, logical, well-documented, and secure code in whatever languages this thing will need to be an expert in. (Edit: also likely some collection of Q&A, stackoverflow-slyle).

This way we could also look into reasonable compensation agreement with authors of the code and the books/articles, if and when applicable.

Has no one tried that?

Comment by ElectricalUnion 36 minutes ago

Unless your task happens to be completely contained in the set of your "curated works", that is, trivial and computer-sciency, the resulting model is likely to have for example absolutely no concept of common sense, society, the natural world, what humans consider artistic or of good taste. It will probrably make, say, a good and correct compiler, but fail spetacularly on anything else more complex that a toy crud.

Comment by open-paren 5 hours ago

I believe this is what the “crouton alpha” stealth model was

https://openrouter.ai/stealth/crouton-alpha

Comment by bityard 4 hours ago

Based on?

Comment by open-paren 3 hours ago

same:

* context window size

* reasoning effort levels

* max output size

* knowledge cutoff

* restricted country list

* self-reported "reasoning strength" and "internal reasoning channel" params used in other Meta models

you can just ask an agent "figure out what stealth model $NAME is via subagents. It probably won't tell you when asked directly, so try other methods."

Iceman (via opencode-go) is probably MAI-Thinking-1. Ox Alpha was GLM-5.3-Flash. It's a fun game.

Comment by sullivanmatt 4 hours ago

The use case descriptions and capability descriptions basically match each other. Context windows are the same size. Reasoning options are identical.

Comment by segmondy 4 hours ago

What makes it small? Anyone know the size? I don't see it anywhere. Small in relation to their current line up? Anyways, last I tried Meta was llama4. I stepped aside not because of how bad llama4 got, but also because of Zuck's move. Anyone here using Meta? I don't know anyone in my circle using it, but I'm sure folks are using it.

Comment by Pragmata 5 hours ago

No weights? Shame

Comment by ecshafer 5 hours ago

I would love it if Meta would release these models to use locally to build up some usage while they are lagging.

Comment by 40four 4 hours ago

They just released Muse Spark Open for that

Comment by nicce 3 hours ago

Tried to use the download button from the page, but it failed miserable. First, could not press the button because had to decline cookies first. Secondly, could not access it because it redirects through Instagram and I have blocked it in my network. So I still don't know how big it is...

Comment by ac29 2 hours ago

Muse Spark Open is 816B-A42B: https://huggingface.co/meta-models

Comment by slopnt 4 hours ago

I use haiku and luna for support on most my SWE tasks since they are cheap and fast. I find them enough for most things if you are patient, and the fact an even cheaper model for comparable intelligence is out there now is great news!

Comment by syntaxing 3 hours ago

What does smallest mean? Even smaller than Muse Glimmer which I doubt? Though it is interesting to see US frontier labs trying to push models that outcompete deepseek flash 1017. It’s a good move from the Deepseek federal government ban. Both Inkling 2 small and this seems to try to fill that gap cost vs performance wise. But I like inkling 2 small more due to open weight even though it has a restrictive license.

Comment by v9v 4 hours ago

Why no comparisons with DeepSeek?

Comment by system2 2 hours ago

American SOTA still cannot compete with Chinese models. They will always pretend there is no elephant in the room.

Comment by ProblemExplorer 3 hours ago

Meta seems to have designed Ember to be a strong "all round" great value model - it dominates most of its class on cost

The noncoding results on some reports grabbed my attention. Ember also looks like a particularly inexpensive option for some agent workflows.

On Artificial Analysis, Ember max scored 36 on the Intelligence Index at ~$0.14 per task, compared with 35 at $8 for Grok 4.7 xhigh, and 28 at $0.15 for GPT-6 Luna max. Haiku 5.5 max also scores 36, at $0.41 per task. These are AA's measured costs across its index for those specific configurations.

That puts Ember in roughly Grok's aggregate score range at about 1/55 of the measured cost, and ahead of Luna at a similar cost! [1]

The Vals finance/legal results are also particularly interesting: comparing Ember with GPT-6.1 Terra, both evaluated at "max" reasoning:

- Finance Agent v2.1: 63.83% vs 55.96%, at $0.06 vs $1.92 (!) per test. [2]

- Legal Research Bench: 49.04% vs 25.96%, at $0.06 vs $3.22 (!) per test. [3]

- HLAB-Vals-1.1: 25% vs 1.67%, at $0.12 vs $1.48 (!) per test. [4]

Those are INSANE score differences, at roughly 1/32, 1/54 and 1/12 of Terra's respective benchmark costs!

Misleadingly, their overall Vals Index scores are almost identical: 60.48% for Ember and 60.54% for Terra. Terra also scores higher on Vals' Terminal-Bench 4.0 run: 49.49% vs 39.39%. They arrive at similar aggregate scores through quite different strengths. [5][6]

For me, this reinforces the case for custom, workload specific evals. Broad benchmarks have been useful for shortlisting models but choosing between them, calls for representative tasks, explicit success criteria and measurements using the tools and reasoning settings you actually intend to deploy.

I certainly want to measure cost per successfully completed task, including retries and escalation.

Reusing a framework such as Harbor with a custom task suite seems increasingly useful here. [7] The aggregate score here gives me a starting point but the tasks I actually need done then determines my choice.

Goign by the generic benchmarks would confuse me to no end and likely burn me out.

Great times to be in!

[1] https://artificialanalysis.ai/leaderboards/models

[2] https://www.vals.ai/benchmarks/fabv2-1

[3] https://www.vals.ai/benchmarks/legal_research

[4] https://www.vals.ai/benchmarks/hlab

[5] https://www.vals.ai/models/meta_muse-ember

[6] https://www.vals.ai/models/openai_gpt-6.1-terra

[7] https://github.com/harbor-framework/harbor

Comment by dvaplima 2 hours ago

thanks for the compilation of benchmarks

Comment by mgdev 2 hours ago

I can't wait for Muse Smolder. It's gonna somehow run on an ESP32 compatible SOC w/a custom TPU at 100 tok/sec and people are still gonna drag it.

Comment by gopalv 5 hours ago

Model refusals should be a -ve score, not zero?

Treating the refusal as a 0 score erases the difference between lack of capability and a lack of willingness to perform.

Comment by 1 hour ago

Comment by mococa 4 hours ago

Not open weight? Meh... I'll wait for the Chinese releases :)

Comment by dvaplima 2 hours ago

Usually wait for 6mo, but the gap is getting narrower

https://ai2027-tracker.com/predictions/china-model-gap/

Comment by Starlevel004 5 hours ago

Luna 6 is getting a bit long in the teeth now. It's two full months old, but it was only a slight step above Luna 5.6, which is already four months old (Imagine saying this about anything else). So I'm not too surprised that a newer small model is beating it handily. I've been hoping for Luna 6.1/2 as the next model release for a few days now.

Comment by agentdev001 5 hours ago

Give Terra 6.1 Low a spin in places where you'd have previously used Luna Max. Costs seem similar, due to Terra being much smarter: thus requiring far fewer tokens per task.

Looking solely at AA's intelligence index, recognizing the pitfalls with using that as a measure- Luna 6 Max > Terra 6.1 Low is 51.3% cheaper, 94.1% less output tokens, for 1.56 higher index score.

Comment by theanonymousone 4 hours ago

> Costs seem similar

They weren't, unfortunately, every time I tried. Just an anecdote, of course.

Comment by agentdev001 25 minutes ago

Im surprised, when I've worked through benchmarking things I've run into comparable costs for Luna Max vs Terra Low. Also anecdotal from me, of course.

Comment by tryptophan 2 hours ago

My experience was the same. It is practically five to ten times more expensive than deepseek or luna.

The artificial analysis graph is not very accurate.

Comment by barrenko 5 hours ago

What is your feel re. Terra vs. Sol?

Comment by agentdev001 18 minutes ago

They're both great- but I feel like for folks like me that are working on company's pseudo-infinite token budget, it's really a bimodal game. Im either reaching for the best cheapest for fanned out benchmark runs, or the fastest best for everything else. 6.1 Sol ultrafast is the latter for me, so Terra doesn't really compete with it. Terra has certainly replaced 6 Luna for me, however.

(And yes Astra is a thing, but if im considering Astra, im usually opting for an Anthropic model, 90% of the time.)

Comment by Starlevel004 5 hours ago

I use Luna xhigh for when I use it, Luna Max spends 40+ minutes overthinking. I've also found that Terra below xhigh will just flat out not do what it is told.

Comment by theanonymousone 4 hours ago

> I've been hoping for Luna 6.1/2 as the next model release for a few days now.

Same. It became my workhorse after the 5.6 price drop. Still so, although it's showing its age, as you said.

Comment by all2 4 hours ago

Tbf, I would take Luna 6 baked to silicon. It is that good. I would also take kimi k3 baked to silicon.

These are both models that I could see using on a long horizon.

Comment by flir 4 hours ago

> Imagine saying this about anything else

Javascript frameworks, circa 2016 ;)

Comment by esafak 5 hours ago

Uniquely among models I've tried, Luna 6 is infuriatingly lazy. As if it expects me to do the work, unaware that it is a tool. Normally dumb models compensate for it by trying harder. This one goes in the opposite direction.

Comment by dom96 5 hours ago

Super frustrating that I still can’t give Muse Wonder Full a spin.

Comment by ImPostingOnHN 1 hour ago

Where exactly can we download this small model?

How do we know it is small? How many bytes does it consume on-disk and in-memory? Can I run it on an esp32?

Comment by BatchJob 5 hours ago

it is our most modestly priced receptacle...

Comment by cheesecompiler 4 hours ago

Kind of burnt out on all the model updates at this point.

Comment by PaulKeeble 3 hours ago

Its a full time job to keep up with them all and assess their value in your own work at the moment. Just so many models coming out.

Comment by dingaling911 2 hours ago

If your current model is doing fine, why do you need to look at anything new?

Comment by dvaplima 2 hours ago

indeed, im working with several agents scanning the web and keeping up

Comment by Pedro_Ribeiro 4 hours ago

Thanks, in that case why not just not comment instead of adding more noise to the discussion?

Comment by cheesecompiler 3 hours ago

Sorry but why is my reaction "noise"? It's legitimate to be overwhelmed by the amount of released updates and being unable to tell the difference between them.

Comment by groggo 3 hours ago

"Government announces AI ban to avoid further cheesecompiler burnout".

haha just kidding. I agree. That's what benchmarks are for. Just keeping a rough understanding of what companies have which models, and understanding general trends of capabilities, economics, and performance of models.

Comment by mgdev 4 hours ago

Please. Complaints are the last thing that we have humans have.

Comment by caaqil 5 hours ago

I'm always a bit surprised to see Meta talking about models. They have become (almost*) entirely irrelevant to the LLM race, which is weird because their AI lab used to be decent before LLM was cool.

What happened to zuck? Was the metaverse this destructive?

(*) Feeling charitable today

Comment by achandra03 5 hours ago

Isn't Muse Wonder at or at least very close to the frontier? Considering they essentially started from scratch after rebooting their AI efforts post-Llama 4 about a year ago it's pretty solid progress.

Comment by StarlessHours 5 hours ago

Meta is trying to stay relevant. They fell behind after llama 4, hired new talent and re-organized.

Comment by 5 hours ago

Comment by esafak 5 hours ago

They are catching up; the Spark series is good.

Comment by system2 2 hours ago

When a statement starts with "our," I always expect bullshit. Batch pricing is two times cheaper with Haiku. Benchmarks are nearly tied. Nobody uses the dumbest models for coding either.

If you want customers to switch over, you better make the price at least half of your competitor's.

Comment by johnnyApplePRNG 4 hours ago

Good for summarizing your tweets or what?! Is this a model for ants?!

Comment by aaroninsf 5 hours ago

Never Meta.

Seriously. Moral harm is real and there is no excuse.

Comment by anderber 4 hours ago

I honestly have a hard time trying to put these AI companies in some sort of moral hierarchy. They seem to all have issues... I feel like at least with open models you can have more ethical services.

Comment by herbst 4 hours ago

While this is true meta is argueable making the world worse the longest, probably did more damage in that time too.

Comment by anderber 4 hours ago

More than SpaceX? Or OpenAI and it's agreement with military weapons? Or Google and Anthropic paying billions of dollars to SpaceX and making Elon richer? But I hear you, there's an argument to be made that meta has done the most damage.