Same. Specifically I was considering Backblaze for our company’s backups (both products, computers and their bucket for server backups. That is no longer the case as of the news.
It should be, but it isn’t. Hence the reason “why can’t you just do what we asked” can often be followed by “then we will find someone who will” in the end.
Pushback is valuable until it becomes obstinance.
If we all somehow had their same crystal ball to know for certain that “stupid shower ideas” won’t work because a specific developer thinks they are bad, there wouldn’t be much need for R&D ever again. I suspect this developer doesn’t have one either, or I’d certainly like to buy it.
Especially when Google is in the far better position to come out ahead…imo.
Edit: so as not to simply spout an opinion, the reasoning I believe this is that Google has a real business already and were already deep into ML and AI research long before they had competitors — they just botched making it a product in the beginning. Anthropic and OpenAI meanwhile are paying hand over fist to subsidize user acquisition. Also, “Deepmind”. I don’t think much more needs to be said regarding that team, and Google has been working on AI since before either Altman or Amodei applied to go to college. They have a vast amount of researchers and resources, their own hardware and data centers (already, not “planned”) and it appears to be showing more recently (in my opinion).
I agree with you in certain circumstances, but not really for internal user inference. OpenRouter is great if you need to maintain uptime, but for basic usage (chat/coding/self-agents) you can do all of what you mentioned and more with a LiteLLM instance. The number of companies that send a bill is rarely a concern when it comes to “is work getting done”, but I agree with you that minimizing user friction is best.
For general use, I personally don’t see much justification as to why I would want to pay a per-token fee just to not create a few accounts with my trusted providers and add them to an instance for users. It is transparent to users beyond them having a single internal API key (or multiple if you want to track specific app usage) for all the models they have access to, with limits and logging. They wouldn’t even need to know what provider is hosting the model and the underlying provider could be swapped without users knowing.
It is certainly easier to pay a fee per token on a small scale and not have to run an instance, so less technical users could definitely find advantage in just sticking with OpenRouter.
1. The LLM provider doesn't know it's you (unless you have personally identifiable information in your queries). If N people are accessing GPT-5.x using OpenRouter, OpenAI can't distinguish the people. It doesn't know if 1 person made all those requests, or N.
> It doesn't know if 1 person made all those requests, or N.
FWIW this is highly unlikely to be true.
It's true that the upstream provider won't know it's _you_ per se, but most LLM providers strongly encourage proxies like OpenRouter to distinguish between downstream clients for security and performance reasons.
Fair point. Would be good to hear from OpenRouter folks on how they handle the safety identifier.
For prompt caching, they already say they permit it, and do not consider it "logging" (i.e. if you have zero retention turned on, it will still go to providers who do prompt caching).
OpenRouter tells you if they submit with your user ID or anonymously if you hover over one of the icons on the provider, eg OpenAI has "OpenRouter submits API requests to this provider with an anonymous user ID.", Azure OpenAI on the other hand has "OpenRouter submits API requests to this provider anonymously.".
But does "anonymous user ID" mean that they make a user ID for you, and it's sticky? If I make a request today and another tomorrow, the same anonymous user ID is sent each time? Or do they keep changing it?
I believe they are static user ids that only OpenRouter knows is you (the anonymous part. Static id is required for any cached pricing. If the user id changes each request, it would be a massive security hole to reuse that cache between requests with different user ids.
Without caching, it would make sense to be per-request (more like a transaction-id, and would make sense to be) as this could then be tied internally back to a user while maintaining external anonymity, but unfortunately I don’t believe that is the case.
1 - I can’t speak to whether that is the case with OpenRouter. However, I suspect that there is more than enough fingerprint and uniqueness inherent to the requests that an AI could probably do a fairly accurate job of reconstructing “possible” sources, even with such anonymity. The result is the same, all your information is still tied to OpenRouter in order to track the billing. That also ignores that OpenRouter is also privy to all that same information. In the end, it comes down to how much you trust your partners.
As for LiteLLM, the company you would pay for inference is going to know it is “you” — the account — but LiteLLM would also have the same effect of appearing to be a single source to that provider. That said, a uniqueness for a user may be passed (as is often with OpenRouter also) for security. Only you know who the users are, that never has to leave your network if you don’t want.
2 - well, you select the providers, so that’s pretty much on you? :-) basically, you are establishing accounts with the inference providers you trust. Bedrock has ZDR, SOC, HIPPA, etc available, even for token inference, as an example. Cost is higher without cache, but you can’t have true ZDR and Cache (that I know of), because a cache would have to be stored between requests. The closest you could get there is maybe a secure inference container but that piles on the cost. Still, plenty of providers with ZDR policies.
LiteLLM is effectively just a proxy for whatever supported (or OpenAI, Anthropic, etc compatible api provider) you choose.
One additional major benefit of OpenRouter is that there is no rate limiting. This is the primary reason why we went with OpenRouter because of the tight rate limiting with the native providers.
Wouldn't they be using the Azure inference API or AWS bedrock on their own accounts and NOT be going through the openAI/Anthropics servers anyways? I just always assumed this is how the big inference "resellers" (openrouter, cursor, etc) were operating.
A lot of inference providers for open models only accept prepaid payments, and managing multiple of those accounts is kind of cumbersome. I could limit myself to a smaller set of providers, but then I'm probably overpaying by more than the 5.5% fee
If you're only using flagship model providers then openrouter's value add is a lot more limited
I haven't noticed any problems with large context requests through OR to e.g. Opus (other than the rate at which my budget gets spent!). Is this a performance thing?
Does OpenRouter perform better than LiteLLM on integration though? I found using Anthropic's models through a LiteLLM-laundered OpenAI-style API to perform noticably worse than using Anthropic's API directly. So I've scrapped considering LiteLLM as an option. It's also just a buggy mess from trying to use their MCP server. The errors it puts out are meaningless, and the UI behaves oddly even in the happy path (error message colored green with Success: prepended).
But if OpenRouter does better (even though it's the same sort of API layer) maybe it's worth it?
OpenRouter performs much, much better than LiteLLM proxy.
In my experience, if OpenRouter offers a model, the API will be supported. They also often have inference providers available that will perform much better than the default provider. Just as an example, Z.ai is sitting at around 10 token/s for GLM 5.1 while friendly is doing 70 token/s for the same model through OpenRouter.
LiteLLM proxy also adds quite some overhead as well.
I have personally settled on a mix of Bifrost as my router which connects to OpenRouter or some other providers that I deem more privacy friendly.
Are you suggesting these researchers somehow have wisdom and aren’t just guessing, and that everyone else are children too naive to understand the technology? It certainly sounds that way from the description you are attempting to apply.
This is two parents disagreeing on whether their child will automatically grow up to be a psychopath with one parent constantly remarking “if you teach that child how to cut bread, they will stab everyone later. If you teach that child to drive, they will run over everyone later”, not the “parents know better” situation you describe.
Think of the stupidest product you can think of and you likely only know about it because people buy/bought them en masse. AI is no different from any other product; plenty will pay/adopt for exactly the reasons you said. There is powerful motivations for people to feel “ahead” of others (or more informed, or more “cool”, or more knowledgeable, or more experienced, or whatever their ego requires), even if their situation is exactly the same.
That said, I’m not sure I follow your statement of less resistance to the development of internal tools when the opposite seems to be the case; companies (or more specifically developers) are perhaps too quick to think they can just vibe-code a replacement for any vendor in a weekend these days.
I’m equally disheartened by the people who dismiss job losses as unlikely because “AI can’t automate entire jobs”.
What do you suggest happens when you automate half of 10 people’s jobs? Do you expect they want to pay for 10 people to operate at 50%, or would someone be more likely to just keep 5 people to do the part they couldn’t automate (yet)? Do you think CEOs will want to add 5 more heads back later, or do you think they will add the minimum necessary and still seek a cheap alternative to fill remaining gaps?
I absolutely agree with you that it COULD make people’s jobs easier, but unless that directly translates into revenue for the company, a relaxing easy day of work isn’t generally the goal of profitable companies.
(Belatedly) yes. Kind of a big argument to grapple with, but let's start by considering everything. I mean, all the stuff, the abstract stuff, that's out there objectively in the universe and in the future, waiting to be discovered. I believe there's quite a considerable amount of it. It's all potentially of interest to us eventually, and only a teeny tiny part of it is comprehensible to us now. That part is at the leading edge, the cutting edge of our enquiries, and in order for us to see and comprehend and even care about that part, it has to relate to us. It has to be oriented to us and our thoughts and things we can use.
You see what I'm getting at? Humans don't really like abstract things. Mathematicians seem to, but I doubt that even mathematics truly has an objective abstract quality that's distant from human concerns. I reckon humans do human mathematics, and it probably has fashions, too, it's probably modern and current, that is, of its time and place.
So you could accept that, but still claim that music relates strongly to mathematics as we know it. Of course there's such a thing as the mathematics of music. I could dispute the value of that to the quality of the music, as being too abstract and niche compared to the evocative qualities of music, where it evokes things in our physical world: the sounds of hitting things with sticks, heartbeats, tones of voice, meaningful instruments such as bugles evoking battles, mazy noodling around evoking contemplative thoughts (is that abstract?) ... but either way, the point is that we live in a sort of parochial Bag End, if Middle Earth represents everything abstractly possible, and so we only understand hobbit things and only appreciate hobbit art. So to speak.
Cool, but isn’t this “the next time”, after “the next time”, after “the next time” already? These companies have been threatened, sued, and incentivized in numerous ways over many years, which has yet to be successful, and yet it seems like you are suggesting “just one more time” will be the impetus for change…this time…you swear…probably?
Note: I don’t disagree or agree, rather, I’m pointing out how flawed the logic is that just one more time will be what it takes.
> isn’t this “the next time”, after “the next time”, after “the next time” already?
No, it’s the time that it worked. The cable company upgraded. That’s all that matters. Whether it’s happened many times or not is irrelevant. The next time will come next time.
> which has yet to be successful
OP said they laid the fiber. It was literally successful. Preëmptively striking your service provider because they might screw you in the future is silly.
reply