Hey everyone! Yagil the founder of LM Studio here. If you want to take Bionic for a spin with GLM 5.2 / Kimi K2.6 / Kimi Coder K2.7, email your lmstudio.ai username to hn-jul16@lmstudio.ai and I'll load your account with some credits!
Try it out for coding (in a "Code" project) and document creation / manipulation (in a "Work" project). In Work projects we have automatic checkpointing for every change the agent makes. Would love to hear your feedback.
sebmellen 20 hours ago [-]
Yagil, thanks for the credits! Lots of fun to try running this locally. I don't understand the negativity in some of the other comments.
This is one of the better agent harnesses I've seen for inspecting reasoning chains, which is super useful for me. Sometimes reading the reasoning is better for my needs than reading the response. I appreciate how transparent that is in contrast with Claude Code/Codex/etc.
One question — I saw you mentioned negotiating ZDR with your "providers" — are you hosting the models yourselves, or is another entity hosting them? If another entity, which group?
13 hours ago [-]
hypercube33 9 hours ago [-]
I love LM studio and this is super cool and I plan on trying it out tonight. I'm absolutely dying for an android link app and I'm jealous of iPhone users
ra 20 hours ago [-]
Can I use my z.ai coding plan API key?
Havoc 12 hours ago [-]
It’s a standard OAI api so technically should work if the software has byok but would break ZAI ToS. They only allow the tools spelled out on their site
They do update that but with lag
lrae 14 hours ago [-]
Wouldn't mind taking a look :)
Imustaskforhelp 20 hours ago [-]
Thanks! I have sent a mail :-D
opallimetea 18 hours ago [-]
[flagged]
sebmellen 18 hours ago [-]
Are you using a local model or one of the hosted ones? I don't think a harness can influence the range of outcomes that much.
secretslol 10 hours ago [-]
I really like LM Studio, but this direction of cloud and trust. I think you've totally blown it with a lot of people, because now they'll be forced to remember another million times they've heard a company say the same kind of thing; 'Zero Data Retention' and 'never training on your data'.
I had a quick browse through your site, but I could not see where you are based, but I get whiffs of more American 'trust me, bro' exports which I've been eye rolling at for two decades now.
inventor7777 19 hours ago [-]
I have never previously tried a agentic harness for local models yet, but I really love LM Studio so I gave Bionic a shot immediately after reading this!
First impression: it works great. I use Codex as my main agent, and the UI looks similar enough that it's familiar and simple to get started. I just pointed it to my existing LM Studio models library, ran Qwen3.6 35B, and the results are exactly what I would hope for.
I did notice some rough edges that might be worth improving, however:
- Current working directory is not the clearest on the main page of the app. It shows the project name, but is missing the prominent working directory label like Codex has.
- The model seems to load when you hit Enter, but it shows "Working" instead of of "Loading model".
- There doesn't seem to be a way to pre load the model, it seems like you have to send it something to load the model.
- I don't see a way to easily unload the model like the eject button in LM Studio without quitting the app
- I pointed it to a directory called "GitHub & Projects" and it somehow ended up making a new folder called "GitHub & Projects". Yes, I know the name is weird but it shouldn't have done that.
gehsty 23 hours ago [-]
This kind of thing just makes me think Apple will get to a point where they have good enough local models and good enough harnesses for doing things, and most normal people will just use them… Does the LLM become another interface to computing?
ibero 23 hours ago [-]
i agree.
i believe that for most people on the street, for most tasks, a Chat GPT 3.5 era LLM is sufficient enough. sprinkle in tool calling and other things, and that becomes enough. if you can prioritize that level of a model on-device (baking it in etc), then you can bifurcate AI users between those unwilling to pay and those who are willing to pay A LOT for frontier model performance.
fy20 3 hours ago [-]
Gemma 4 is already closer to gpt-4o level of intelligence, however it's knowledge is much lower.
trvz 13 hours ago [-]
Anyone even a little bit serious will be paying either way. If not for the cloud models, then for hardware upgrades.
SOLAR_FIELDS 22 hours ago [-]
This question hinges on whether model advancement plateaus enough for machine sized models to compare to frontier performance. If it does, the answer is yes. If it doesn’t, the answer is no
jfaat 15 hours ago [-]
I disagree. The point of the frontier models is to do everything as well as possible ("AGI" race or whatever) but smaller models with some RL are going to be the clear winner for a ton of use cases. Think about all the use cases for LLMs that would never be economical at frontier inference costs, and in no way need it. You don't need or even want a phd polymath helping you with small productivity tasks that most people use computers for every day. It's often overwrought and annoying. I don't even really like the frontier models for coding for this reason. They're constantly blowing up scope and you have to fight it constantly.
cptskippy 21 hours ago [-]
More likely, it's going to be whether frontier models advance enough that most people would be willing to pay for them. Right now they don't, but a model you can run locally for free on hardware you already own is very compelling because, while they're not as good as Frontier Models, they're still pretty good.
Tools like Opencode demonstrate that when you box them in tightly enough they can actually be pretty competent.
c7b 21 hours ago [-]
Neural machines were always going to be an alternative computing paradigm to von Neumann machines. Had it not been for Minsky we would arguably have gotten to a point where they're useful sooner. But why do you say that as if it's a small thing?
mark_l_watson 9 hours ago [-]
Apple’s System Model is actually pretty good but is hard-limited to a 4K context length. This is OK for small Python utilities that need a model for applications that operate on small amounts of data, but is a disappointing limitation.
That said, and this is off topic: Siri on the newest iOS beta is surprisingly good now. I asked it what model it was using yesterday and it said Gemini for difficult problems, then secure Apple model in cloud, and local Apple model.
jjice 7 hours ago [-]
Looks like iOS 27 ships with a 20B sparse model for on-device work, so I wonder if that comes with a context length bump. They're still pushing for better local models, which is a good sign for the future for their products.
I have thought this for a while. Computing 1.0 meant that we needed to learn the computer’s language to interact with the computer fully. Computing 2.0 is that now the computer has learned our language instead.
23 hours ago [-]
bigyabai 21 hours ago [-]
Why wait? People are already doing their work on OpenAI and Anthropic's servers, Apple Intelligence servers could quickly subsume any "local" model work that you want to do.
That way everyone has access, even with older devices, and it's a subscription! Then Apple can tie their APIs into the ecosystem you love at a flat cost you can afford. No need to support local model integration in the first place, problem solved.
chickensong 19 hours ago [-]
> Does the LLM become another interface to computing?
It already is.
gehsty 3 hours ago [-]
It literally is, but people don’t use LLMs to interact with their computers they use touch screens or mouse and a keyboard.
This is why I think it will be the platform owners who end up winning - if iPhone / Mac / iPad ship with a good enough local model that can directly interact with the OS and apps and the web people might actually start talking to their computers.
dominotw 22 hours ago [-]
or the other way where primary interfaces ppl use computing arent apple devices like laptops and phones.
doctorpangloss 19 hours ago [-]
based on their apple intelligence demos they are optimizing their products for their core demo of 55-95 year old boomers who talk out loud to think and read every page of the nytimes. you are miles away from the US consumer product experience here.
gehsty 3 hours ago [-]
You say this like it a bad thing my 90yr old grandmother can use WhatsApp and iCloud Photos - if they can get that demographic using LLMs meaningfully, it probably means they’ve cracked the interface.
marcus_holmes 19 hours ago [-]
Boomers were born up until 1965, so the youngest boomer is 61 this year.
Please don't slur us older GenX as boomers.
13 hours ago [-]
satvikpendem 21 hours ago [-]
Why would I use this over any other harness? I suppose they're wrapping it all up in a nice package for enterprise, especially ones that want to control their LLM usage for cost and data security compared to the cloud frontier models.
slopinthebag 20 hours ago [-]
There’s actually not a lot of good harnesses that aren’t slopped together python or JS codebases, which are actually model-agnostic and don’t do some really silly things like bloating the context, too much compaction etc.
There is no way I’m running a python or JS agent which is probably vibe coded, it’s just too much risk from a security and supply chain standpoint.
Geezus_42 20 hours ago [-]
The harness is a security threat because it's vibecoded but the code you're having agents write for you isn't? Lol
ubermon 12 minutes ago [-]
I think it is more on distrusting python/typescript in general. Those dynamic scripting language are too easy to poison.
slopinthebag 17 hours ago [-]
I don’t vibe code though, to the extent I actually use agents to generate code it all gets reviewed and it’s essentially exactly what I would have written myself. It’s mostly things like mechanical refactors, boilerplate generation, etc.
Geezus_42 3 minutes ago [-]
Don't you think that is how the majority of vibecoders see themselves?
https://github.com/pjlsergeant/byre can add small but very convenient layer around whichever agent you want, initially trapping it in a specific folder, but you can easily mount more, firewall it, bring in MCPs and skills, or whatever. It's unambiguously AI-assisted, but a very very great deal of thought and work has gone in to the design of it. Ask your favourite agent to review the code if you're skeptical.
IronWolve 16 hours ago [-]
It was fun to try out which local models made better agents. Few thoughts.
* locked to a single dir, so no system wide access.
* no local web search, can be fixed ddg or local mcp.
* no ssh, I want to have it ssh into my server and do the
work.
* doesnt show the model being loaded, needs a bar/%
counter.
* Can you drag/drop documents in the work dirs, or only + add them?
I love lm-studio, so cant wait to see how this goes. For local I normally use opencode + lmstudio.
esoltys 5 hours ago [-]
[dead]
codazoda 21 hours ago [-]
I’m worried about the switch in business model here, which is part of the reason I just switched to LM Studio from Ollama.
> use the largest frontier open source models through LM Studio Secure Cloud
satvikpendem 21 hours ago [-]
Use Unsloth Studio, it's actually open source and I trust Unsloth via their quantized models a lot more than LM Studio.
docheinestages 21 hours ago [-]
Any VC-backed company will eventually pivot.
satvikpendem 16 hours ago [-]
Use now then switch later. Such is the life of a software user.
jckahn 21 hours ago [-]
And also enshittify
SamInTheShell 21 hours ago [-]
At least code isn't a moat anymore. Have a weekend and want your own Discord? Doable.
Geezus_42 20 hours ago [-]
Prove it.
SV_BubbleTime 17 hours ago [-]
Well, it’ll LOOK like discord!
Geezus_42 46 minutes ago [-]
Right! Lol
trvz 13 hours ago [-]
To be fair, Ollama has been pushing their own cloud offering just as much.
For some of the latest big models (from Minimax, GLM, Kimi, etc.) even after months they still don't provide official downloadable versions, just their cloud one.
Havoc 12 hours ago [-]
Indeed though Ollama has had drama from the start while LM studio has had zero to my knowledge. So bit more faith in them pulling off the transition in a mostly graceful manner
thehamkercat 23 hours ago [-]
A friendly reminder that both LM Studio app and now this new LM Studio Bionic app are closed source.
Since most people are unaware of this fact.
nodja 22 hours ago [-]
Yup, it's the main reason I don't use LM studio more. I only use it to try out new models/quants, then use llama.cpp directly to host them. LM Studio also doesn't do stuff like audio input and often has bugs that pure llama.cpp doesn't so it can be a net negative for certain use cases.
satvikpendem 21 hours ago [-]
Unsloth Studio is open source, run by the same Unsloth that produces some of the best quants in the business, I'd advise people to switch.
maxloh 22 hours ago [-]
Yeah, we already have many open-source agent systems. If you prefer a UI, OpenCode itself has a beta desktop app.
I don't think we need closed-source developer tools, especially ones where they might restrict access if they decide to start charging for them later.
cpursley 22 hours ago [-]
Does anyone know about their stack - is it a native app? It's fairly well designed for what it is in terms of desktop apps.
woadwarrior01 22 hours ago [-]
The desktop app's GUI is electron.js based.
simondotau 19 hours ago [-]
Perfect for when you have unified memory and you want to use as much of it as possible for running the LLM. </sarcasm>
minimaxir 23 hours ago [-]
And? Is that a scandal?
thehamkercat 23 hours ago [-]
Not a scandal ofc, but people use local models mostly for privacy
and using a closed-source, VC-backed app that might change anything in the next update might not be best for privacy
SOLAR_FIELDS 22 hours ago [-]
It also seems weird to be closed source and then market yourself as a tool designed for things that are open
some people want to make money, others want to improve social progress.
Happy to clarify which is who and who is which.
radial_symmetry 23 hours ago [-]
You can do both
blitzar 21 hours ago [-]
Sure you can ... simply start by taking donations to benefit all mankind and then once you have done enough of that go private, ipo and join the tres commas club.
folkrav 22 hours ago [-]
Sure. I still don't think it's particularly controversial to acknowledge that the two don't necessarily align either, and that neither really incentivizes the other.
Less unanimous and debatable, but many would say they more often do not align than the opposite.
freehorse 22 hours ago [-]
You can do both, until you can't, and then usually making money trumps social progress.
watwut 23 hours ago [-]
No one involved in LLMs want to improve social progress. That is simply not part of the equasion.
simondotau 19 hours ago [-]
Social progress is literally the story of labor saving devices. Creating more with less labor means a society of relative abundance, with more time for education, creativity, leisure and personal growth. Don't be fooled into thinking LLMs are any different to the tractor, the loom, the calculator, or shoe factories in the third world.
dofm 9 hours ago [-]
Don't be fooled into thinking that the tractor or the loom brought about positive social progress.
It wasn't the machines that caused the social progress. It was the organised labour movements (and the variously destructive or violent sabotage movements) pushing back against the machines that made the social progress.
The tractor and the loom made life noticeably worse for the people they replaced until they organised against them, and even then it took decades for them to get back the limited kind of quality of life (and health) an agricultural worker might have had.
I do find it interesting how many AI people make tractor and loom analogies without the slightest grasp of the true history of the industrial revolution.
Any social progress that comes from AI will not be something AI people can claim as their success, because it will have happened at their expense. There will be pushback; it may be violent.
simondotau 8 hours ago [-]
The suggestion that violence may be necessary does not strengthen your argument, it exposes its weakness. A movement that cannot defend its ideals without romanticising destructive force is not protecting social progress. It is to fundamentally misunderstand where social progress comes from, or what represents a good life.
As for the rest, it's a remarkably selective account of the industrial revolution. Your argument relies upon the unspoken inference that pre-industrial agricultural life was somehow a better life that was diminished by progress.
The tractor did not merely replace farm workers. It allowed a tiny fraction of the population to produce vastly more food while releasing millions of people from exhausting agricultural labour. Perhaps you would prefer a life of hard physical labour, low output, poor medicine, food insecurity, and little protection against injury, disease, or harvest failure. Perhaps you don't like that you toil away in an air-conditioned office writing video game code, for a salary that you wish was higher, because you dread the cost of your next smartphone purchase.
Technological change created enormous productivity gains. Yes, these often unfairly disadvantage workers trained in the outgoing industries. But I don't hear you arguing for a return to hand-weaving, because the societal benefit of progress is inarguable.
dofm 7 hours ago [-]
> The suggestion that violence may be necessary does not strengthen your argument, it exposes its weakness.
I am not at all suggesting it is necessary; you are projecting that onto me. I am merely observing that it might happen. Already is happening.
> Your argument relies upon the unspoken inference that pre-industrial agricultural life was somehow a better life that was diminished by progress.
Well for several generations life expectancy actually fell, working hours increased, general health worsened, child and infant mortality worsened. It took until the 1900s for those measures to improve. Child labour became an outsized evil until work and education reforms happened really only 140 years ago.
Do you not understand that industrial work was not less exhausting until workers rights movements did something about it? Not at all less dangerous? Not at all less dirty, less violent, better paid, more free, more secure?
For the first century or so, the industrial revolution (and the second agricultural revolution) made the average worker's life worse — less stable, less constant, fewer rights, a destruction of traditions and social cohesion.
The social progress you argue that technology drove, it drove as largely a negative influence. Did we get modern medicine after two hundred years or so? Of course. Was there industrial progress? Of course. But social progress happened as a result of the labour organisation of people to stop their lives being made radically worse by a class of new extremely rich capitalists. Productivity gains are not social progress, nor are they a proxy for it.
> Perhaps you don't like that you toil away in an air-conditioned office writing video game code, for a salary that you wish was higher, because you dread the cost of your next smartphone purchase.
This is childish straw man projection of which you should be a bit ashamed, IMO. That is the kindest thing I can say of it.
simondotau 6 hours ago [-]
Call it whatever you like. I'm perfectly fine with my use of obvious hyperbole, and I have no plans to reconsider when I'm unconvinced that the misreading of it was sincere.
Claiming that life expectancy "fell for several generations" is a poor reading of the data. Most generously there was a period of stagnation predominantly caused by a combination of scarlet fever and rapid urbanisation.
More generally you're arguing against positions I haven't espoused and which I don't think even represent areas of disagreement. Productivity gains and social organisation are not opposing forces. The surplus from higher output made shorter hours, better pay, public health, and medicine sustainable, making it possible for labour movements to secure larger shares and better terms of that surplus. We live in an amazing world now, if you take a moment to stop and think about what actually matters.
watwut 13 hours ago [-]
> Social progress is literally the story of labor saving devices.
Historically it was not. They are two mostly unrelated things. Creating more with less labor can mean the richest getting richer while middle class becomes poor and poor become hopelessly poor.
> Creating more with less labor means a society of relative abundance, with more time for education, creativity, leisure and personal growth.
It never meant that, this is fairly tale. We got 40 hours workweek rather then 80 hours workweek because of political movements and fights, not because of technology. Labor saving device, on itself, leads to two outcomes: you work as much as before, but produce more (cue all the burned out overworked ai coders) or you get unemployed desperately looking for new work.
simondotau 11 hours ago [-]
Political movements influenced how abundance was shared. They did not create that abundance.
Production and consumption are directly linked: societies can consume only what they produce. When individuals become more productive, society has more to consume. And indeed there is more. So much more. What you call a "fairy tale" is, thankfully, the world we live in.
When considering prosperity, it's important to emphasise what's real. So what if Warren Buffett owns a massive number of shares in notionally highly valued corporations? There are plenty of carrots available in the supermarket, and at an extraordinarily low prices by historical standards. Warren Buffett is not personally consuming some grotesquely disproportionate share of the carrots ... or pretty much any commodity which constitutes real material prosperity.
watwut 10 hours ago [-]
Political movements we are talking about were result of rapidly worsening living conditions for large groups of people. They did not necessary followed technological progress, technological progress was not necessary condition for their existence.
Nor were they result of a new abundance.
> When considering prosperity, it's important to emphasise what's real.
Yes. And what did actually happened in the past.
> So what if Warren Buffett owns a massive number of shares in notionally highly valued corporations?
Meanwhile, Warren Buffett : "There’s class warfare, all right, but it’s my class, the rich class, that’s making war, and we’re winning."
The accumulation of power and money by smaller group of people means loss of power by the rest people. Those with power then use it to their own benefit while rest of us dont have that option. This result in the small group having even more power, while needs and wants of the rest of us are ignored. In turn, it means worsening conditions for the people on the loosing side.
And you see this process happening as we speak here. You see it at political decisions being made by the administration, at climate change approach, at unaccountability of the Epstein group of friends, at wars for oil, at decay of the democracy.
Materially you see K shaped economy, in us being unable to buy RAM and electronics, loosing rights, loosing control of devices to the corporations, basic necessities getting more expensive and so on.
simondotau 9 hours ago [-]
That is a pile of unrelated grievances pretending to be an economic argument. Once again, your fixation on politics obscures the central point: politics can redistribute production; it cannot redistribute what has not been produced.
The RAM example is particularly absurd. Two decades ago, ordinary consumers were not complaining about the price of 64 or 128 GB of RAM because owning that much was barely imaginable. You are complaining about access to a product that exists only because capitalism drove extraordinary technological progress. Obsessing over RAM shows a lack of historical perspective and a badly distorted idea of what constitutes a good life.
Yes, concentrated wealth means Warren Buffett has a disproportionately large share of influence. But that does not mean Warren Buffett also consumes a disproportionately large share of the resources that represent human flourishing. Ownership, political power, and physical consumption are different things. Corporate shares are not sandwiches ripped from the mouths of children.
And blaming every disliked event, from wars to Epstein to electronics prices, on conspiracies involving the ultra-wealthy is exactly the kind of unhinged, zero-sum populism that corrodes democracy. Instead of evidence and analysis, we're given a third rate morality play about a wicked elite stealing everything from "the people". Extreme polarisation and populist thinking are corrosive forces in modern society.
speedgoose 23 hours ago [-]
No, but you are a lot more vulnerable to enshittification and you don’t have the advantages of open source applications.
It’s an important criteria to have in mind when you select an application.
felooboolooomba 23 hours ago [-]
I'm interested in your thought process. How did you get from his initial statement opening of "A friendly reminder ..." to thinking that this was a scandal?
minimaxir 23 hours ago [-]
It's a common discussion trope to imply malfeasance in response to good news, which is a way to non-constructively shut down a conversation particularly without elaboration. In this particular case I legit didn't understand what the OP was actually implying because they did not elaborate.
NamlchakKhandro 23 hours ago [-]
Yes.
Is that a problem for you comrad?
cicko 23 hours ago [-]
It is, cowboy
yogthos 16 hours ago [-]
it sure is bootlicker
dofm 21 hours ago [-]
I am aware of it, and I dabble with Unsloth Studio and use the llama-server approach.
I would obviously prefer an open source, open weights stack.
But I guess a paradox is that as long as there are open source options I could use, a solid agentic environment that I can use with my own open weights is something I might pay for, in a similar sort of way to paying for a Mac when I could use only Linux.
If someone wanted to make their entire income from, say, making the BBEdit of LLM harnesses, that would be a viable strategy. Sooner or later people need to make an income somewhere. My own feeling is that Apple should acquire LM Studio, but if they said "this is $X per year" I might consider it, given the attention to detail.
Aerialoo 15 hours ago [-]
I seriously don't understand the apple glazing. They are pretty much completely out of the picture and are one of the most anti consumer anti open-source companies. Yet developers STILL hope for Apple acquisitions and being saved... Or something like that...
We have llama.cpp, vLLM, opencode, hermes and countless other mature great open source tools that don't require a anti-consumer company's hands in the jar.
dofm 11 hours ago [-]
OK. Not sure I agree Opencode is mature, having used it enough and seen the source code, but if that is the level of attention to detail that now qualifies as “mature” then I understand why my comment upset people so much and I apologise.
I think it ought to be possible to understand my comment without assuming it is Apple fandom; it’s not.
I am just one of those developers who use Linux daily as a remote/VM work environment, from a MacBook, from personal choice. I have for decades. There is a kubuntu laptop on my desk too (a Framework) which is fine; I have more or less always had GUI linux around too. But my own LLM dabbling is happening on an M1 Max, like so many others, because I am not a gamer or home build PC guy (this century at least) so I don’t have a PC with a GPU. (Yet. I am LLM-cynical but I can see running a dedicated machine for this in principle)
The suggestion about Apple (who I don’t believe are yet “out of the picture” in any meaningful sense) acquiring LM Studio was an idle reflection on them being the most Apple-ish of the options, that’s all. They need something; LM Studio is closest to what they need.
Lio 13 hours ago [-]
I would guess it's because Apple are a hardware vendor that at least talks about privacy as a selling point.
The M series chips provide large amounts of GPU RAM so I think people can see a world where Apple sells local first, privacy focused, inference based on hardware sales.
LM Studio's strategy seems to fit that very well.
bel8 11 hours ago [-]
I would respect Apple's privacy talk if they facilitated running Linux. But it takes years of reverse engineering to get it barely useful on a Mac.
So they put revenue above true privacy, as companies do. But they are the loudest to spit the bs while selling black boxes.
Lio 10 hours ago [-]
Sorry I think you misunderstand me.
I'm not pitching Apple as the best or only solution, I'm saying that selling local inference seems to fit with their current marketing goals.
Selling high end, high margin computers that have lots available GPU RAM with at least lip service to local inference for privacy fits with what they already do.
I don't see anything to stop say Framework making the same pitch but with Linux but since Framework don't design their own chips they don't have quite so many levers to pull as Apple do.
lacoolj 5 hours ago [-]
It says this is a separate app than regular LM Studio but I'm not seeing a way to download it.
Great to see LM Studio expanding into agent workflows. Local model tooling keeps getting better, and having an open-source option for this is valuable for developers who want to keep their data private.
HugoDz 9 hours ago [-]
Is this open source or not ?
metalliqaz 5 hours ago [-]
Very exciting, I'll be trying this out soon.
I wish LM Studio played better with AMD hardware. It would be really great to have an off-the-shelf solution that "just works" on Radeon.
jFriedensreich 9 hours ago [-]
What the hell, as far as i see this is not open source but has the tag line "Meet LM Studio Bionic, the AI agent made for open models" How can someone justify this complete clash of values, who is even an audience for something this. cursed, the agent and harness layer being open is probably more important than your models as you can just switch models every week and they are commodity.
blitzar 23 hours ago [-]
I am not sure I get this. It seems on first glance like just another harness ...
woadwarrior01 22 hours ago [-]
Ultimately, the onus at every VC backed local LLM startup is to launch a cloud based offering, because that's the only potential path in sight for venture scale returns.
freehorse 22 hours ago [-]
For now, it seems that direction that lm studio is taking for enterprise market is "local ai deployment support". They recently launched "lm link" which basically uses tailscale to create e2ee connections between computers running lm studio where you are logged in. Granted one can also setup tailscale or their own vpn themselves and use llama-server, but I guess it is simpler to provide it out of the box. In any case I am not sure pivoting from running local models to "cloud offering" (as in providing llm inference at their severs) is a sensible choice granted there is already competition in that space and they have no leverage there. The highest expected path imo would be to be bought by a company that makes (esp open weights) llms and has a similar business plan around enterprise contracts with local deployments.
woadwarrior01 19 hours ago [-]
> In any case I am not sure pivoting from running local models to "cloud offering" (as in providing llm inference at their severs) is a sensible choice granted there is already competition in that space and they have no leverage there.
I agree. Incidentally, this is exactly what ollama are doing too.
christoph 15 hours ago [-]
I can definitely see a world where you run stuff local first, for all the reasons we know. Sometimes, you are going to want more powerful models, faster, you’re travelling, etc. You might only use these 5 to 10% of the time, but I’m guessing that’s the market they want.
freehorse 12 hours ago [-]
But one can already switch models in a harness between llm providers. I doubt such a pivot can get some much traction.
Normal_gaussian 23 hours ago [-]
Built to work with lmstudio, one of the leading easy to use local model servers. LMStudio is the closest to plug-and-play without sacrificing play that I've seen; a harness that works well with it is nothing to sniff at. Its not earth shattering either.
cjonas 23 hours ago [-]
Wouldn't most opensource harnesses work with lmstudio? I assume it has an "openai" style chat API like every other model provider? What's special about it vs langchain deep agents or pi or pydantic-ai?
solarkraft 21 hours ago [-]
Yes. I don’t see it either. It looks like a competent app (converging on the same principles as others) but what they are advertising as differentiators simply isn’t relevant to its purpose.
blitzar 23 hours ago [-]
I guess it lives or dies by the harness quality then - on open models run locally by plug and players and models that fit onto peoples laptops - that is going to be quite the handicap to overcome.
I run lmstudio personally with a range of harnesses (open and closed) and can't say there is that much of a leap to getting everything talking https://lmstudio.ai/docs/integrations
solarkraft 21 hours ago [-]
Maybe they optimized it really hard for small models. That would be impressive, but hardly a good selling argument. I want a harness that can do both and OpenCode looks just fine for that purpose (and it also has everything else I need).
To me this looks like another case of bundling things that shouldn’t be bundled (the harness with the UI) making both worse off because you can’t individually focus on each component. They’ve done this before, bundling a decent UI with decent inference for IMO no good reason, combining the downsides of each instead of letting people mix+match.
belowavgiq 22 hours ago [-]
built to work with an OpenAI API compatible endpoint, just like any other harness...
and if someone can't figure out how to write down an address it's very likely they also can't figure out how to make local models not suck for coding, and would likely switch back to codex/cc after 15 minutes anyways.
tokai 22 hours ago [-]
How does it work with lmstudio? Its a separate standalone application.
minimaxir 23 hours ago [-]
If you want to use local models, it's more ergonomic than fussing with GGUFs or using LM Studio as a server host and setting up the link to an agent yourself. Although, the model selector is the same as with LM Studio itself which can be overwhelming if you don't know what to look for.
solarkraft 21 hours ago [-]
It’s a harness + UI with really confusing messaging. Made for open models ... why specifically? Zero data retention ... well duh, it’s a client????
slopinthebag 20 hours ago [-]
Maybe I missed it, but can we use this with other cloud api’s? Like using Deepseek directly from their platform via their OpenAI endpoints?
Lalabadie 8 hours ago [-]
I looked for the same, and no, not currently. It's only local models or LM Studio's own API.
slopinthebag 5 hours ago [-]
Boo. Ah well.
eliaskg 12 hours ago [-]
I was searching for that option as well but didn't find it. I guess it's just a new UI + harness for LMStudio in the end.
crowcountry 9 hours ago [-]
Does this support ACP?
fishfasell 21 hours ago [-]
It says no data retention or training on your data, but I assume that doesn't hold true for the frontier cloud models you connect to?
yags 21 hours ago [-]
It does! We negotiated ZDR with our providers. We consider that a condition to make things available to our users: in this case cloud inference and web search/extract.
(I’m the founder of LM Studio)
jedisct1 9 hours ago [-]
https://swival.dev is an AI agent that was designed for open models since day 1.
ekianjo 11 hours ago [-]
trusting them to not retain your data and never train on it is for fools. this is something they say to gain adoption and that they will reverse with an obscure 'privacy policy update' down the road.
zuzululu 13 hours ago [-]
another cool project that i will never use because frontier models are just too far ahead
Capitanai 10 hours ago [-]
[flagged]
elhart05 11 hours ago [-]
[flagged]
1g10k 8 hours ago [-]
[flagged]
przemarzec 11 hours ago [-]
[dead]
Fr4nZ82 22 hours ago [-]
[flagged]
jkwang 13 hours ago [-]
[dead]
nttylock 21 hours ago [-]
[flagged]
ShipVoicedev 23 hours ago [-]
[flagged]
Rendered at 20:36:20 GMT+0000 (Coordinated Universal Time) with Vercel.
Try it out for coding (in a "Code" project) and document creation / manipulation (in a "Work" project). In Work projects we have automatic checkpointing for every change the agent makes. Would love to hear your feedback.
This is one of the better agent harnesses I've seen for inspecting reasoning chains, which is super useful for me. Sometimes reading the reasoning is better for my needs than reading the response. I appreciate how transparent that is in contrast with Claude Code/Codex/etc.
One question — I saw you mentioned negotiating ZDR with your "providers" — are you hosting the models yourselves, or is another entity hosting them? If another entity, which group?
They do update that but with lag
I had a quick browse through your site, but I could not see where you are based, but I get whiffs of more American 'trust me, bro' exports which I've been eye rolling at for two decades now.
First impression: it works great. I use Codex as my main agent, and the UI looks similar enough that it's familiar and simple to get started. I just pointed it to my existing LM Studio models library, ran Qwen3.6 35B, and the results are exactly what I would hope for.
I did notice some rough edges that might be worth improving, however:
- Current working directory is not the clearest on the main page of the app. It shows the project name, but is missing the prominent working directory label like Codex has. - The model seems to load when you hit Enter, but it shows "Working" instead of of "Loading model". - There doesn't seem to be a way to pre load the model, it seems like you have to send it something to load the model. - I don't see a way to easily unload the model like the eject button in LM Studio without quitting the app - I pointed it to a directory called "GitHub & Projects" and it somehow ended up making a new folder called "GitHub & Projects". Yes, I know the name is weird but it shouldn't have done that.
i believe that for most people on the street, for most tasks, a Chat GPT 3.5 era LLM is sufficient enough. sprinkle in tool calling and other things, and that becomes enough. if you can prioritize that level of a model on-device (baking it in etc), then you can bifurcate AI users between those unwilling to pay and those who are willing to pay A LOT for frontier model performance.
Tools like Opencode demonstrate that when you box them in tightly enough they can actually be pretty competent.
That said, and this is off topic: Siri on the newest iOS beta is surprisingly good now. I asked it what model it was using yesterday and it said Gemini for difficult problems, then secure Apple model in cloud, and local Apple model.
https://runaihome.com/blog/wwdc-2026-apple-ai-home-lab-verdi...
That way everyone has access, even with older devices, and it's a subscription! Then Apple can tie their APIs into the ecosystem you love at a flat cost you can afford. No need to support local model integration in the first place, problem solved.
It already is.
This is why I think it will be the platform owners who end up winning - if iPhone / Mac / iPad ship with a good enough local model that can directly interact with the OS and apps and the web people might actually start talking to their computers.
Please don't slur us older GenX as boomers.
There is no way I’m running a python or JS agent which is probably vibe coded, it’s just too much risk from a security and supply chain standpoint.
* locked to a single dir, so no system wide access.
* no local web search, can be fixed ddg or local mcp.
* no ssh, I want to have it ssh into my server and do the work.
* doesnt show the model being loaded, needs a bar/% counter.
* Can you drag/drop documents in the work dirs, or only + add them?
I love lm-studio, so cant wait to see how this goes. For local I normally use opencode + lmstudio.
> use the largest frontier open source models through LM Studio Secure Cloud
For some of the latest big models (from Minimax, GLM, Kimi, etc.) even after months they still don't provide official downloadable versions, just their cloud one.
Since most people are unaware of this fact.
I don't think we need closed-source developer tools, especially ones where they might restrict access if they decide to start charging for them later.
and using a closed-source, VC-backed app that might change anything in the next update might not be best for privacy
https://lmstudio.ai/app-privacy
Happy to clarify which is who and who is which.
Less unanimous and debatable, but many would say they more often do not align than the opposite.
It wasn't the machines that caused the social progress. It was the organised labour movements (and the variously destructive or violent sabotage movements) pushing back against the machines that made the social progress.
The tractor and the loom made life noticeably worse for the people they replaced until they organised against them, and even then it took decades for them to get back the limited kind of quality of life (and health) an agricultural worker might have had.
I do find it interesting how many AI people make tractor and loom analogies without the slightest grasp of the true history of the industrial revolution.
Any social progress that comes from AI will not be something AI people can claim as their success, because it will have happened at their expense. There will be pushback; it may be violent.
As for the rest, it's a remarkably selective account of the industrial revolution. Your argument relies upon the unspoken inference that pre-industrial agricultural life was somehow a better life that was diminished by progress.
The tractor did not merely replace farm workers. It allowed a tiny fraction of the population to produce vastly more food while releasing millions of people from exhausting agricultural labour. Perhaps you would prefer a life of hard physical labour, low output, poor medicine, food insecurity, and little protection against injury, disease, or harvest failure. Perhaps you don't like that you toil away in an air-conditioned office writing video game code, for a salary that you wish was higher, because you dread the cost of your next smartphone purchase.
Technological change created enormous productivity gains. Yes, these often unfairly disadvantage workers trained in the outgoing industries. But I don't hear you arguing for a return to hand-weaving, because the societal benefit of progress is inarguable.
I am not at all suggesting it is necessary; you are projecting that onto me. I am merely observing that it might happen. Already is happening.
> Your argument relies upon the unspoken inference that pre-industrial agricultural life was somehow a better life that was diminished by progress.
Well for several generations life expectancy actually fell, working hours increased, general health worsened, child and infant mortality worsened. It took until the 1900s for those measures to improve. Child labour became an outsized evil until work and education reforms happened really only 140 years ago.
Do you not understand that industrial work was not less exhausting until workers rights movements did something about it? Not at all less dangerous? Not at all less dirty, less violent, better paid, more free, more secure?
For the first century or so, the industrial revolution (and the second agricultural revolution) made the average worker's life worse — less stable, less constant, fewer rights, a destruction of traditions and social cohesion.
The social progress you argue that technology drove, it drove as largely a negative influence. Did we get modern medicine after two hundred years or so? Of course. Was there industrial progress? Of course. But social progress happened as a result of the labour organisation of people to stop their lives being made radically worse by a class of new extremely rich capitalists. Productivity gains are not social progress, nor are they a proxy for it.
> Perhaps you don't like that you toil away in an air-conditioned office writing video game code, for a salary that you wish was higher, because you dread the cost of your next smartphone purchase.
This is childish straw man projection of which you should be a bit ashamed, IMO. That is the kindest thing I can say of it.
Claiming that life expectancy "fell for several generations" is a poor reading of the data. Most generously there was a period of stagnation predominantly caused by a combination of scarlet fever and rapid urbanisation.
More generally you're arguing against positions I haven't espoused and which I don't think even represent areas of disagreement. Productivity gains and social organisation are not opposing forces. The surplus from higher output made shorter hours, better pay, public health, and medicine sustainable, making it possible for labour movements to secure larger shares and better terms of that surplus. We live in an amazing world now, if you take a moment to stop and think about what actually matters.
Historically it was not. They are two mostly unrelated things. Creating more with less labor can mean the richest getting richer while middle class becomes poor and poor become hopelessly poor.
> Creating more with less labor means a society of relative abundance, with more time for education, creativity, leisure and personal growth.
It never meant that, this is fairly tale. We got 40 hours workweek rather then 80 hours workweek because of political movements and fights, not because of technology. Labor saving device, on itself, leads to two outcomes: you work as much as before, but produce more (cue all the burned out overworked ai coders) or you get unemployed desperately looking for new work.
Production and consumption are directly linked: societies can consume only what they produce. When individuals become more productive, society has more to consume. And indeed there is more. So much more. What you call a "fairy tale" is, thankfully, the world we live in.
When considering prosperity, it's important to emphasise what's real. So what if Warren Buffett owns a massive number of shares in notionally highly valued corporations? There are plenty of carrots available in the supermarket, and at an extraordinarily low prices by historical standards. Warren Buffett is not personally consuming some grotesquely disproportionate share of the carrots ... or pretty much any commodity which constitutes real material prosperity.
Nor were they result of a new abundance.
> When considering prosperity, it's important to emphasise what's real.
Yes. And what did actually happened in the past.
> So what if Warren Buffett owns a massive number of shares in notionally highly valued corporations?
Meanwhile, Warren Buffett : "There’s class warfare, all right, but it’s my class, the rich class, that’s making war, and we’re winning."
The accumulation of power and money by smaller group of people means loss of power by the rest people. Those with power then use it to their own benefit while rest of us dont have that option. This result in the small group having even more power, while needs and wants of the rest of us are ignored. In turn, it means worsening conditions for the people on the loosing side.
And you see this process happening as we speak here. You see it at political decisions being made by the administration, at climate change approach, at unaccountability of the Epstein group of friends, at wars for oil, at decay of the democracy.
Materially you see K shaped economy, in us being unable to buy RAM and electronics, loosing rights, loosing control of devices to the corporations, basic necessities getting more expensive and so on.
The RAM example is particularly absurd. Two decades ago, ordinary consumers were not complaining about the price of 64 or 128 GB of RAM because owning that much was barely imaginable. You are complaining about access to a product that exists only because capitalism drove extraordinary technological progress. Obsessing over RAM shows a lack of historical perspective and a badly distorted idea of what constitutes a good life.
Yes, concentrated wealth means Warren Buffett has a disproportionately large share of influence. But that does not mean Warren Buffett also consumes a disproportionately large share of the resources that represent human flourishing. Ownership, political power, and physical consumption are different things. Corporate shares are not sandwiches ripped from the mouths of children.
And blaming every disliked event, from wars to Epstein to electronics prices, on conspiracies involving the ultra-wealthy is exactly the kind of unhinged, zero-sum populism that corrodes democracy. Instead of evidence and analysis, we're given a third rate morality play about a wicked elite stealing everything from "the people". Extreme polarisation and populist thinking are corrosive forces in modern society.
It’s an important criteria to have in mind when you select an application.
Is that a problem for you comrad?
I would obviously prefer an open source, open weights stack.
But I guess a paradox is that as long as there are open source options I could use, a solid agentic environment that I can use with my own open weights is something I might pay for, in a similar sort of way to paying for a Mac when I could use only Linux.
If someone wanted to make their entire income from, say, making the BBEdit of LLM harnesses, that would be a viable strategy. Sooner or later people need to make an income somewhere. My own feeling is that Apple should acquire LM Studio, but if they said "this is $X per year" I might consider it, given the attention to detail.
I think it ought to be possible to understand my comment without assuming it is Apple fandom; it’s not.
(see e.g. https://news.ycombinator.com/item?id=48805487 for how I feel about brand loyalty)
I am just one of those developers who use Linux daily as a remote/VM work environment, from a MacBook, from personal choice. I have for decades. There is a kubuntu laptop on my desk too (a Framework) which is fine; I have more or less always had GUI linux around too. But my own LLM dabbling is happening on an M1 Max, like so many others, because I am not a gamer or home build PC guy (this century at least) so I don’t have a PC with a GPU. (Yet. I am LLM-cynical but I can see running a dedicated machine for this in principle)
The suggestion about Apple (who I don’t believe are yet “out of the picture” in any meaningful sense) acquiring LM Studio was an idle reflection on them being the most Apple-ish of the options, that’s all. They need something; LM Studio is closest to what they need.
The M series chips provide large amounts of GPU RAM so I think people can see a world where Apple sells local first, privacy focused, inference based on hardware sales.
LM Studio's strategy seems to fit that very well.
So they put revenue above true privacy, as companies do. But they are the loudest to spit the bs while selling black boxes.
I'm not pitching Apple as the best or only solution, I'm saying that selling local inference seems to fit with their current marketing goals.
Selling high end, high margin computers that have lots available GPU RAM with at least lip service to local inference for privacy fits with what they already do.
I don't see anything to stop say Framework making the same pitch but with Linux but since Framework don't design their own chips they don't have quite so many levers to pull as Apple do.
Am I blind?
The word "Bionic" is never mentioned on the homepage
Any other ideas?
I wish LM Studio played better with AMD hardware. It would be really great to have an off-the-shelf solution that "just works" on Radeon.
I agree. Incidentally, this is exactly what ollama are doing too.
I run lmstudio personally with a range of harnesses (open and closed) and can't say there is that much of a leap to getting everything talking https://lmstudio.ai/docs/integrations
To me this looks like another case of bundling things that shouldn’t be bundled (the harness with the UI) making both worse off because you can’t individually focus on each component. They’ve done this before, bundling a decent UI with decent inference for IMO no good reason, combining the downsides of each instead of letting people mix+match.
and if someone can't figure out how to write down an address it's very likely they also can't figure out how to make local models not suck for coding, and would likely switch back to codex/cc after 15 minutes anyways.
(I’m the founder of LM Studio)