Hacker Newsnew | past | comments | ask | show | jobs | submit | more decide1000's commentslogin

I use it on a 24gb gpu Tesla P40. Very happy with the result.


Out of interest, roughly how many tokens per second do you get on that?


Like 4. Definitely single digit. The P40s are slow af


P40 has memory bandwidth of 346GB/s which means it should be able to do around 14+ t/s running a 24 GB model+context.


Not sure why I got downvoted - literally the first result (for me) says the best result[0] is 11t/s at Q3. Everything else is single digits, like 2-8t/s. Also considering that its not supported anymore[1] (It's Compute Capability is 6.1, not supported by cuda anymore) and it's power draw, I'd highly recommend anyone interested in ML stay far away from it - even if its all you can afford.

While the memory bandwidth is decent, you do actually need to do matmuls and other compute operations for LLMs, which again its pretty slow at

[0]: https://old.reddit.com/r/LocalLLaMA/comments/1dcdit2/p40_ben... [1]: https://developer.nvidia.com/cuda-gpus


Pretty cool, but where is the Android app?


Coming very soon, it's been a long process with the Android store. Thank god for react native and expo


Can't wait to try would love to test this - can you share an apk?


Email ishaan@omnara.com with your google play store email account. He can invite you to being a tester for the android app while it's making its way through the review process!


About the Webapp; better search and filter on previous conversations. Filters on model type. Better errors when context is too big. Forking conversations would be nice. Better export options. Copy whole convo (not just response or reply).

On the LLM: It's too positive. I don't always want it to follow my ideas and I don't want to hear how much my feedback is appreciated. Act like a machine. Also the safety controls are too sensitive sometimes. Rlly annoying because there is no way to continue the conversation. I like gpt4.5 because i can edit the canvas. Would like to have that with all models.

Also some stats like sentiment and fact check would be nice. Because it gives nuances in answers I want to see with the stats how far from the truth or bias I am.

And the writing.. Exaggerating, too many words, spelling mistakes in European languages.


This is great! I hear you on the overly positive responses. You mention "act like a machine", but is there perhaps a desire/need for a more human-feeling interface?


If I talk to a business relation or friend, they give me an answer with a few words or a few sentences. Asking a question to an LLM gives me paragraphs. I am not sure if a different interface is going to change that.

I use LLMs now by asking questions, copy/pasting stuff around. A human-feeling in terface sounds more like TTS and I don't think that is human for me.


Qwen3-30A-A3B-2507 is much faster on my machine compared to gpt-oss-20B. This leaderboard does not reflect that.


The model is good and runs fine but if you want to be blown away again try Qwen3-30A-A3B-2507. It's 6gb bigger but the response is comparable or better and much faster to run. Gpt-oss-20B gives me 6 tok/sec while Qwen3 gives me 37 tok/sec. Qwen3 is not a reasoning model tho.


It was fun because it was open. Now it's just another brand seeking dollars.


Ollama at its core will always be open. Not all users have the computer to run models locally, and it is only fair if we provide GPUs that cost us money and let the users who optionally want it to pay for it.


I think it’s the logical move to ensure Ollama can continue to fund development. I think you will probably end up having to add more tiers or some way for users to buy more credits/gpu time. See anthropic’s recent move with Claude code due to the usage of a number of 24/7 users.


I’m not throwing the towel on Ollama yet. They do need dollars to operate, but still provide excellent software for running models locally and without paying them a dime.


^ this. As a developer, Ollama has been my go-to for serving offline models. I then use cloudflare tunnels to make them available where I need them.


Although it is open, its really just all code borrowed from llama.cpp.

If you want to see where the actual developers do the actual hard work, go use llama.cpp instead.


C'mon CF. What are you doing? You are literally breaking the internet with your police behaviour. Starts to look like the Great Firewall.


Not affiliated with CF in any way. Respectfully disagree. Calling out bad actors is in the public interest.


CF is a bad actor. They ruin internet. They own more and more parts of it.


It's not in my interest that a tech company from the US decides what a bad actor is.


Why does it take so long? Existing EU recipes are already compliant Kraft’s European products have for years used natural colours such as turmeric, paprika, beet juice or no colouring at all. That is why the 2025 U.S. pledge to go dye free by 2027 is largely irrelevant on this side of the Atlantic. So 2027? That does not make sense at all.. it's a n economic perspective, not a healthy one.


Supply chains.

EU and US supply chains are vastly different, plus shifting the production lines from one to another doesn't happen overnight. This means that it could well take two years to fully move all their production facilities off synthetic food dyes.


Tumeric sometimes contains lead. I think only in India so far, but the FDA is about to move lots of testing to the states. Hopefully on a roadtrip or layover you won't have to research each state you are in before eating.


Demanding an entire industry change everything overnight doesn't work. Suppliers have to ramp up production, processes have to be reworked, purchasing contracts have already been set a year in advance.


I don't understand why Americans accept this behavior from corporates. They are basically poisoning people for economic reasons. Why don't they use that extra profit, made over the health of millions, to speed up this process.


Because the USA politicians for the most part do what corporate interests want. When it looked like consumer interests might gain a foothold in 1960-1970s, a member of the US chamber of commerce wrote the Powell memo as a guide to corporate responses to consumer activism, this is still followed by USA corporations and he was appointed to US Supreme Court to put his influence on legal standards for generations. He’s not the first nor the last as traditionally in the USA due to how weak the framers made the federal government originally, a private public partnership is always exalted as the best of both worlds for governance but this leads to regulatory capture we have now without sufficient safeguards. edit: an adjacent comment mentions GRAS,Generally Recognized as Safe https://en.wikipedia.org/wiki/Generally_recognized_as_safe - laws in the USA last forever unless new laws supercede them so this law set in stone thousands of chemicals (including the aforementioned dyes) as safe because they didn't kill anyone quickly in 1958 and companies are free to use those in food - only a handful of items have been removed since then because the FDA is required to scientifically prove harm, unlike how those items got put on the list.


Guessing it's to ramp-up suppliers, change equipment over, and stockpile enough for the transition.


I understand the need to phase out/in ingredients in this situation, however I've never understood when there is a simple ban on an unnecessary ingredient why it takes long. I'm specifically talking about those "microbeads" in bodywash that were banned a few years ago. The companies got years to phase them out. They served no real purpose and were not replaced with anything. Companies just had to stop adding them to the bodywash - why give them years to do so? I get that labelling would be inaccurate so give them a few months to change that.


Of course the beads served a purpose: they were abrasive and exfoliating. And they were given time because they have to sell their existing inventory and use all the beads they already have purchased to put in their products.


Not sure why you're greyed out, because you're correct that those were literally exfoliant face washes. (No real loss in phasing them out because there are much better ways to exfoliate than rubbing your face with little pellets, but it wasn't some meaningless design choice.)


Sorry, yes I realize their purpose, but I meant the product still worked without them, so they could have stopped adding them and continue to sell the product. I guess the same is technically true for food dyes. I just mean that there are times where the ingredient is critical to the product, microbeads was not one of them.


With something like that it's tough, though, because they sold those as an _exfoliating wash_ product, but if you take away the beads it's not exfoliating anymore. So you'd either have to rebrand it to remove any mentions of exfoliation (at which point it becomes a totally different product) or rework the ingredients list to include a "chemical" exfoliant rather than the "physical" exfoliant (but that also drastically alters the product, especially since some people avoid chemical exfoliants).

Not defending microbeads—those products were truly shit, both for your skin and for the environment—I just want to illustrate that it might not be so straightforward.


I think we give these companies too much leeway. I live in Canada and there is a palpable hate on for American's following the "51st State" talk. This has led to increased patriotism and companies are cashing in. Within weeks, many retail products started to relabel and emphasize "Made In Canada". So it can be done, we shouldn't listen to them when they tell us "it's not that easy" because they've demonstrated it is easy when it means potentially more revenue.


And run out existing contracts with existing suppliers.


They'd have to scrap all the food currently in the production and distribution pipeline, plus there would be a gap in food delivery as producers switch over to a new process. It's less disruptive to transition gradually.

Similar to why the USAID closure was gradual and gave aid recipients plenty of time to find new donors, because we wouldn't want hundreds of thousands of women and children to die of starvation and disease just to save a few bucks or wring out more viral memes [1].

[1] https://www.nytimes.com/2025/06/22/us/politics/usaid-cuts-do... / https://archive.is/5BIAF


Logitech mx master s2 and as keyboard Logitech k800


This is insane. They aimed his corporate account. Europeans move away from the US even faster now. This guy is literally breaking decades old relationships.


You think Europe would really behave any different?

Israel is the wedge and leverage to eliminating governments of Iran, Pakistan, China and then India and weakening Russia further.

Colonialism hasn't gone anywhere, evidence? Europe fully protects settlers and their ambitions despite what they say publicly. It is a long road but the most realistic one they have.


> Israel is the wedge and leverage to eliminating governments of Iran, Pakistan, China and then India and weakening Russia further.

China and India? and Russia? Israel doesn't care about them.

Israel is using America and Europe to gain control of the middle east, so they want to eliminate Iran (and just like with Iraq, they want the US to do it for them).

Pakistan is somewhat friendly so not really a threat to Israel's control over the middle east.

US and Europe do not benefit from this. Europe actually suffers as instability in middle east causes a refugee crisis in Europe.


> You think Europe would really behave any different?

And do you think America would run their government on the software of European tech monopolies?


They run their tech and production on Chinese monopolies


Europe probably would behave differently to be honest, but we’re so far away from either of us finding out if we’re correct or not that its not worth talking about.

What we should instead focus on is digital sovereignty.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: