Article: https://proton.me/blog/deepseek
Calls it “Deepsneak”, failing to make it clear that the reason people love Deepseek is that you can download and it run it securely on any of your own private devices or servers - unlike most of the competing SOTA AIs.
I can’t speak for Proton, but the last couple weeks are showing some very clear biases coming out.
You must log in or register to comment.
this is obviously talking about their web app, which most people will be using. In this special instance, it was clearly not the LLM itself censoring the Tiananmen Square, but a layer on top.
i have not bothered downloading and asking deepseek about Tiananmen Square. so i cannot know what the model would have generated. however, it is possible that certain biasses are trained into any model.
i am pretty sure, this blog is aimed at the average user. while i wouldn’t trust any LLM company with my data, i certainly wouldn’t want the chinese government to have them. anyone that knows how to use (ollama)[https://github.com/ollama/ollama] should know these telemetry data don’t apply to running locally. but for sure, pointing it out in the blog would help.
@ToxicWaste @JOMusic the censorship is trained into the ollama models too. But of course the self-hosted model cannot send anything to China, so at least the whole tracking issue is avoided.
1978 US Automotive Companies: If we make a product that locks our customers in, they’ll be our customers forever!
1978 Japanese Automotive Companies: The US gave us their required parameters. If we make a product that works then customers will keep buying our stuff.
2025 US Tech Companies: If we make our products contingent on proprietary software and hardware, we’ll lock them in.
2025 Chinese Tech Companies: The US gave us their required parameters. If we make a product that works and they can utilize freely, they’ll keep buying our stuff.
Not our first rodeo.
It’s simple:
bad.Well you just made me choke on my laughter. Well done, well done.
🤣
CHYNA
Jesus fuckin Christ, just marry Trump at this point, Mister proton CEO.
I cancelled my Proton renewal for January and am very happy with Mullvad VPN.
Mozilla VPN runs Mullvad under the hood as well.
Does Mullvad have “Secure Core” option like Proton does? I’m kinda thinking about switching.
I remember mullvad has less servers than proton and I hear they get blacklisted often. Have you encountered anything like this?
I’ve been “blocked by network security” on reddit. Switching to the next server resolves the issue.
Since ditching Proton for Tuta and Mailbox…I haven’t missed anything and I’m saving money.
I got a proton vpn subscription a while ago and they upgraded me to unlimited for the same price. So I think I’m paying like $6.25/month for an unlimited plan. I feel like it’s too good to leave. If I do tuta’s plan that’s $3, then another $4 for simplelogin, and $5 for mullvad. So that’s $12 a month if I leave my plan.
I get it, and please, you do you. There’s no issue.
I’d just add that I can save money using Amazon, but I try to avoid it when I can. I’ll pay a little extra when I can, for the greater good.
You have two email addresses in both Tuta and Mailbox? Any particular reason for that, that you could share with us? 🙏
I have two domains, one in each of Tuta and Mailbox. It was originally so I could try both out, but now I figure it doesn’t hurt to keep 'em separated. I’m still new to non-proton so I am sort of still feeling things out.
Nothing really too interesting or tricky about it, just bred out of curiosity.
Ah I see. So now to the possibly tough question, if you had to choose only one, or recommend only one of them to someone who wants to make a minimal amount of new email addresses, which one would you recommend over the other? 😅 Or maybe a third option?
I think I’d need some more time to really answer, but on the outset, I find Mailbox.org’s interface more intuitive with more settings and generally feels cleaner and more streamlined. Creating aliases and domain aliases in mailbox seems more proton-like in its simplicity.
Tuta I think is more private and secure, but bits of their interface and app need polish. One reason I think Tuta is more secure despite them both touting security and privacy is that Mailbox search works immediately, whereas Tuta requires you to agree to a permission and states it stores everything locally to you so it may take up space. I think Tuta isn’t doing any server-side indexing of any kind? Unsure.
edit: Mailbox doesn’t have a native app, and Tuta has a native app but I think it’s largely a webview. Notifications work OK but you’ll click on a notification and then have to wait for the app to actually connect and resync before you can view it.
he’s probably right. the company wants to be disruptive, and it’s normal for any company to steal data. you can self host the current model, but that doesn’t mean this will always be the case. certainly they will want to make a profit at some point. it’s day 1 silicon valley shit
Goddammit I had such high hopes for Proton. Was planning on that being my post-Google main. Now what. 💀
I’ve been happy with Fastmail for 10 years, though they’re Australian and not European. Might look into a European alternative at some point but so far I’ve had no reason to switch.
Tutanota and Mailfence have a free tier.
At this point I’m this 🤏 close to
hosting my own emailabandoning it all and living in a cabin in yhe woodsMan, I wish self hosted email was a reasonable thing to do. But it’s a pain to set up the server and the domain stuff, and once you do, if anyone ever spammed off that IP, you’re probably screwed anyway because good luck getting off the blacklists.
Anything European-based to recommend? I’d like something as far-removed from America as possible, respecting GDPR, privacy, etc., but with a good-sized free-tier storage. I don’t think I need more than a couple GB for email. Calendar included would be a big plus as well. 😅 Probably asking for a lot here…
honestly probably worth paying for for something if it means enough to you
I use Infomaniak Mail or ikmail for short. They give you 20GB free, have a whole suite (calendar and others), and are Swiss based. It can also link to other mail clients under the free tier. Only hurdle is using a VPN or proxy for initial sign up, but that can be turned off for daily usage.
Tutanota is gdpr but only 1GB free storage. They do offer calendar for free as well with open sourced apps.
Thanks! I saw Tuta from the previous comment and thought 1 GB is a bit on the small side, kind of like Proton. But not too expensive to go up a tier either. 👍
I found this while searching on my own. Might help someone else. 🤷♂️
People got flack for saying Proton is the CIA, Proton is NSA, Proton is a joint five-eyes country intelligence operation despite the convenient timing of their formation and lots of other things.
Maybe they’re not, maybe their CEO is just acting this way.
But consider for a moment if they were. IF they were then all of this would make more sense. The CIA/NSA/etc have a vested interest in discrediting and attacking Chinese technology they have no ability to spy or gather data through. The CIA/NSA could also for example see a point to throwing in publicly with Trump as part of a larger agreed upon push with the tech companies towards reactionary politics, towards what many call fascism or fascism-ish.
My mind is not made up. It’s kind of unknowable. I think they’re suspicious enough to be wary of trusting them but there’s no smoking gun, yet there wasn’t a smoking gun that CryptoAG was a CIA cut-out until some unauthorized leaks nearly a half century after they gained control and use of it. We know they have an interest in subverting encryption, in going fishing among “interesting” targets who might seek to use privacy-conscious services and among dissidents outside the west they may wish to vet and recruit.
True privacy advocates should not be throwing in with the agenda of any regime or bloc, especially those who so trample human and privacy rights as that of the US and co. They should be roundly suspicious of all power.
In other words, honeypot. And an US plant in Switzerland…
To be fair its correct but it’s poor writing to skip the self hosted component. These articles target the company not the model.
DeepSeek is open source, meaning you can modify code(new window) on your own app to create an independent — and more secure — version. This has led some to hope that a more privacy-friendly version of DeepSeek could be developed. However, using DeepSeek in its current form — as it exists today, hosted in China — comes with serious risks for anyone concerned about their most sensitive, private information.
Any model trained or operated on DeepSeek’s servers is still subject to Chinese data laws, meaning that the Chinese government can demand access at any time.
What??? Whoever wrote this sounds like he has 0 understanding of how it works. There is no “more privacy-friendly version” that could be developed, the models are already out and you can run the entire model 100% locally. That’s as privacy-friendly as it gets.
“Any model trained or operated on DeepSeek’s servers are still subject to Chinese data laws”
Operated, yes. Trained, no. The model is MIT licensed, China has nothing on you when you run it yourself. I expect better from a company whose whole business is on privacy.
What??? Whoever wrote this sounds like he has 0 understanding of how it works. There is no “more privacy-friendly version” that could be developed, the models are already out and you can run the entire model 100% locally. That’s as privacy-friendly as it gets.
Unfortunately it is you who have 0 understanding of it. Read my comment below. Tldr: good luck to have the hardware
Obviously you need lots of GPUs to run large deep learning models. I don’t see how that’s a fault of the developers and researchers, it’s just a fact of this technology.
I understand it well. It’s still relevant to mention that you can run the distilled models on consumer hardware if you really care about privacy. 8GB+ VRAM isn’t crazy, especially if you have a ton of unified memory on macbooks or some Windows laptops releasing this year that have 64+GB unified memory. There are also websites re-hosting various versions of Deepseek like Huggingface hosting the 32B model which is good enough for most people.
Instead, the article is written like there is literally no way to use Deepseek privately, which is literally wrong.
as I said in my original comment, it’s not only VRAM that matters.
I honestly doubt that even gamer laptops can run these models with a usable speed, but even if we add up the people who have such a laptop, and those who have a PC powerful enough to run these models, they are tiny fractions of those that use the internet on the world. it is basically not available to those that want to use it. ot is available to some of them, but not nearly all who may want it
So I’ve been interested in running one locally but honestly I’m pretty confused what model I should be using. I have a laptop with a 3070 mobile in it. What model should I be going after?
There are already other providers like Deepinfra offering DeepSeek. So while the the average person (like me) couldn’t run it themselves, they do have alternative options.
Is it Open Source? I cannot find the source code. The official repository https://github.com/deepseek-ai/DeepSeek-R1 only contains images, a PDF file, and links to download the model. But I don’t see any code. What exactly is Open Source here?
I don’t see the source either. Fair cop.
Thanks for confirmation. I made a top level comment too, because this important information gets lost in the comment hierarchy here.
Open source is in general wrong term in all of these “open source” LLM’s (like LLAMA and R1), the model is shared, but there is no real way of reproducing the model. This is because the training data is never shared.
In my mind open source means that you can reproduce the same binary from source. The models are shared for free, but not “open”.
Down votes be damned, you are right to call out the parent they clearly dont articulate their point in a way that confirms they actually understand what is going on and how an open source model can still have privacy implications if the masses use the company’s hosted version.
I think they mean privacy friendly version of the infrastructure could be developed.
To be fair, most people can’t actually self-host Deepseek, but there already are other providers offering API access to it.
There are plenty of step-by-step guides to run Deepseek locally. Hell, someone even had it running on a Raspberry Pi. It seems to be much more efficient than other current alternatives.
That’s about as openly available to self host as you can get without a 1-button installer.
Running R1 locally isn’t realistic. But you can rent a server and run it privately on someone else’s computer. It costs about 10 per hour to run. You can run it on CPU for a little less. You need about 2TB of RAM.
If you want to run it at home, even quantized in 4 bit, you need 20 4090s. And since you can only have 4 per computer for normal desktop mainboards, that’s 5 whole extra computers too, and you need to figure out networking between them. A more realistic setup is probably running it on CPU, with some layers offloaded to 4 GPUs. In that case you’ll need 4 4090s and 512GB of system RAM. Absolutely not cheap or what most people have, but technically still within the top top top end of what you might have on your home computer. And remember this is still the dumb 4 bit configuration.
Edit: I double-checked and 512GB of RAM is unrealistic. In fact anything higher than 192 is unrealistic. (High-end) AM5 mainboards support up to 256GB, but 64GB RAM sticks are much more expensive than 48GB ones. Most people will probably opt for 48GB or lower sticks. You need a Threadripper to be able to use 512GB. Very unlikely for your home computer, but maybe it makes sense with something else you do professionally. In which case you might also have 8 RAM slots. And such a person might then think it’s reasonable to spend 3000 Euro on RAM. If you spent 15K Euro on your home computer, you might be able to run a reduced version of R1 very slowly.
You don’t need that much ram to run this
How much do you need? Show your maths. I looked it up online for my post, and the website said 1747GB, which is completely in-line with other models.
You can run an imitation of the DeepSeek R1 model, but not the actual one unless you literally buy a dozen of whatever NVIDIA’s top GPU is at the moment.
A server grade CPU with a lot of RAM and memory bandwidth would work reasonable well, and cost “only” ~$10k rather than 100k+…
I saw posts about people running it well enough for testing purposes on an NVMe.
Can you link that post?
Thanks!
That’s cool! I’m really interested to know how many tokens per second you can get with a really good U.2. My gut is that it won’t actually be better than the 24VRAM+96RAM cache setup this user already tested with though.
Those are not deepseek R1. They are unrelated models like llama3 from Meta or Qwen from Alibaba “distilled” by deepseek.
This is a common method to smarten a smaller model from a larger one.
Ollama should have never labelled them deepseek:8B/32B. Way too many people misunderstood that.
I’m running deepseek-r1:14b-qwen-distill-fp16 locally and it produces really good results I find. Like yeah it’s a reduced version of the online one, but it’s still far better than anything else I’ve tried running locally.
Have you compared it with the regular qwen? It was sissy very good
The main difference is speed and memory usage. Qwen is a full-sized, high-parameter model while qwen-distill is a smaller model created using knowledge distillation to mimic qwen’s outputs. If you have the resources to run qwen fast then I’d just go with that.
I think you’re confusing the two. I’m talking about the regular qwen before it was finetuned by deep seek, not the regular deepseek
Its so cute when chinese is sprinkled in randomly hehe my little bilingual robot in my pc
The 1.5B/7B/8B/13B/32B/70B models are all officially DeepSeek R1 models, that is what DeepSeek themselves refer to those models as. It is DeepSeek themselves who produced those models and released them to the public and gave them their names. And their names are correct, it is just factually false to say they are not DeepSeek R1 models. They are.
The “R1” in the name means “reasoning version one” because it does not just spit out an answer but reasons through it with an internal monologue. For example, here is a simple query I asked DeepSeek R1 13B:
Me: can all the planets in the solar system fit between the earth and the moon?
DeepSeek: Yes, all eight planets could theoretically be lined up along the line connecting Earth and the Moon without overlapping. The combined length of their diameters (approximately 379,011 km) is slightly less than the average Earth-Moon distance (about 384,400 km), allowing them to fit if placed consecutively with no required spacing.
However, on top of its answer, I can expand an option to see its internal monologue it went through before generating the answer, which you can find the internal monologue here because it’s too long to paste.
What makes these consumer-oriented models different is that that rather than being trained on raw data, they are trained on synthetic data from pre-existing models. That’s what the “Qwen” or “Llama” parts mean in the name. The 7B model is trained on synthetic data produced by Qwen, so it is effectively a compressed version of Qen. However, neither Qwen nor Llama can “reason,” they do not have an internal monologue.
This is why it is just incorrect to claim that something like DeepSeek R1 7B Qwen Distill has no relevance to DeepSeek R1 but is just a Qwen model. If it’s supposedly a Qwen model, why is it that it can do something that Qwen cannot do but only DeepSeek R1 can? It’s because, again, it is a DeepSeek R1 model, they add the R1 reasoning to it during the distillation process as part of its training. (I think they use the original R1 to produce the data related to the internal monologue which it is learns to copy.)
What makes these consumer-oriented models different is that that rather than being trained on raw data, they are trained on synthetic data from pre-existing models. That’s what the “Qwen” or “Llama” parts mean in the name. The 7B model is trained on synthetic data produced by Qwen, so it is effectively a compressed version of Qen. However, neither Qwen nor Llama can “reason,” they do not have an internal monologue.
You got that backwards. They’re other models - qwen or llama - fine-tuned on synthetic data generated by Deepseek-R1. Specifically, reasoning data, so that they can learn some of its reasoning ability.
But the base model - and so the base capability there - is that of the corresponding qwen or llama model. Calling them “Deepseek-R1-something” doesn’t change what they fundamentally are, it’s just marketing.
There is no “fundamentally” here, you are referring to some abstraction that doesn’t exist. The models are modified during the fine-tuning process, and the process trains them to learn to adopt DeepSeek R1’s reasoning technique. You are acting like there is some “essence” underlying the model which is the same between the original Qwen and this model. There isn’t. It is a hybrid and its own thing. There is no such thing as “base capability,” the model is not two separate pieces that can be judged independently. You can only evaluate the model as a whole. Your comment is just incredibly bizarre to respond to because you are referring to non-existent abstractions and not actually speaking of anything concretely real.
The model is neither Qwen nor DeepSeek R1, it is DeepSeek R1 Qwen Distill as the name says. it would be like saying it’s false advertising to say a mule is a hybrid of a donkey and a horse because the “base capabilities” is a donkey and so it has nothing to do with horses, and it’s really just a donkey at the end of the day. The statement is so bizarre I just do not even know how to address it. It is a hybrid, it’s its own distinct third thing that is a hybrid of them both. The model’s capabilities can only be judged as it exists, and its capabilities differ from Qwen and the original DeepSeek R1 as actually scored by various metrics.
Speaking of its “base capabilities” is a meaningless floating abstraction which cannot be empirically measured and doesn’t refer to anything concretely real. It only has its real concrete capabilities, not some hypothetical imagined capabilities.
I suspect the enshittification of proton is fast approaching.
Im stuck. Is there a Guide for a fast approaching full suit switch?
Caleneder, Passwords, Email, Drive?
many services will not accept email addresses that are not Gmail or protonmail or outlook etc. I don’t know what you would use
Email and calendar: Tuta, Posteo, Mailbox.org, Disroot
Passwords: Bitwarden, KeepassXC
Drive: Filen.io
Proton have been too noisy from the very start .
It might be that they’re equating the name with the app and company, not the open source model, based on one of the first lines:
AI chat apps like ChatGPT collect user data, filter responses, and make content moderation decisions that are not always transparent.
Emphasis mine. The rest of the article reads the same way.
Most people aren’t privacy-conscious enough to care who gets what data and who’s building the binaries and web apps, so sounding the alarm is appropriate for people who barely know the difference between AI and AGI.
I get that people are mad at Proton right now (anyone have a link? I’m behind on the recent stuff), but we should ensure we get mad at things that are real, not invent imaginary ones based on contrived contexts.
Most people aren’t privacy-conscious enough to care who gets what data
I assume most people who pay for proton don’t fall into this category
Yeah it’s a fair call, but to me it is the very context of why people are made at Proton that makes me suspicious of articles like this.
I can’t find the original summary post someone made, but here’s the last response from Proton CEO. Read the comments as well to get a good summary: https://www.reddit.com/r/ProtonMail/comments/1i2nz9v/on_politics_and_proton_a_message_from_andy/
TL;DR: Proton used their official accounts to share CEO’s pro-US-Republican thoughts as their official stance. They since apologized and said they would use personal account to share those thoughts. But (IMO) now having posted this blog on the actual Proton website, it says to me that there are some serious bias alignment issues with a company that is supposed to be a safe-haven away from all of that.
Here is a general write up about the CEO showing their maga colors.
More happened in the reddit thread though that added some more elements, like the ceo opting for a new user name with “88” in it (a common right wing reference), his unprompted use of the phrase “didnt mean to trigger you,” him evasively refusing to clarify what his stance actually was because “that would be more politics,” on and on. You can read through that thread here, although proton corporate are mods, so i have no idea what they may have deleted at this point.
The thread was full of “mask on” behavior that is pretty transparent to anyone experienced with the alt right on the internet.
Thank you so much! That was way beyond what I could have hoped.
I’ll read the link you provided in a bit, but that does sound really bad. Must suck to work at a company you think is helping people stay private only to have the CEO come out as pro-fascism.
No problem mate. The thread is a mess, but if you read the comments below the top pinned one, you’ll see most of the salient points that pissed people off. The “color” i mentioned above came from all over that thread, with some ofi t deleted. I know he edited/deleted the “triggered” comment when he was called out, but he never expanded on why he claimed the GOP was the “party of the little guy” and why all the “corporate dems” needed to be thrown out to get anything done. He also opted not to respond at all to people asking why he thought the party of tech billionaires was suddenly going to crack down on tech billionaires besides saying he really liked J.D vance, a tech millionaire whose political run was funded by, get this, tech billionaire Peter theil.
Dude fawned very publicly over tech billionaire maga, who will do clearly do nothing for privacy and monopoly busting, while pretending that the real issue is chuck shumers daughters working in tech.
I wasn’t a customer of theirs (I’m always skeptical of super-popular-anything), but I think I’ll look elsewhere for secure email.
Not because of this article, which I think makes some decent points, but because I would worry in the back of my mind that the Officers of the company would happily bow to their demigods and start secretly tracking people as a show of fealty.
it is certainly that. but recently its become very trendy to hate Proton, so its just easier to do that instead of thinking. I’m really disappointed in this community
There are many llms you can use offline
Including DeepSeek: https://huggingface.co/deepseek-ai
Deepseek works reasonably well, even at cpu only in ollama. I ran the 7b and 1.5b models and it wasn’t awful. 7b slowed down as the convo went on, but the 1.5b model felt pretty passable while I was playing with it
