OpenRouter
107
MENTIONS
43
EPISODES
39
PODCASTS
Search complete. 107 mentions across 43 episodes found for "OpenRouter".
Sep 28, 2026
Sam BadriGUEST
As a sector, what I do is I, I go kind of around the world, and I meet with shareholders of Cisco or prospective shareholders, and they ask me very kind of critical questions like, "How do you know your demand is real? How do you know the, the demand is durable?" Or, or, "What do you guys see?" Or, "What do you have to believe?" I think for AI, because so much of that, so much of the, the tech trend hinges on specifically AI, you need to believe that we are still early from a token consumption perspective.
Sam BadriGUEST
And if I take the token data that's readily available, some of it is on OpenRouter.
Sam BadriGUEST
There, you know, you can go on, go on X. There's hundreds of publications.
Sam BadriGUEST
So much of the, the token kinda charts or frameworks are showing just explosive growth.
Alex MashrabovGUEST
Yeah, look, I think let's just look at the outcomes which we have today.
Alex MashrabovGUEST
Out of all the incumbents in the United States, when I look at open router data, the only company which is relevant is Google out of all the incumbents.
Alex MashrabovGUEST
When I look in China, where probably obsession with benchmarks probably is less, we have Tencent, Xiaomi, Alibaba, three incumbents being completely relevant, and obviously, like, ByteDance obviously trying to catch up as well
Harry StebbingsHOST
What's your takeaway from that?
Alex MashrabovGUEST
So most of the companies in the world today, most of the businesses, they don't necessarily need Astra specifically.
Alex MashrabovGUEST
They don't necessarily need the newest Fable model.
Alex MashrabovGUEST
That's why, like, Open Router reports that share of open source models went from below thirty to over sixty within this year
Harry StebbingsHOST
What do you think share of open models will be in two years' time?
Zach VorhiesGUEST
And so what you see right now is that you see a collapse of, uh, closed source models.
Zach VorhiesGUEST
Well, it was until, you know, cl- Opus 5.5. We'll have to see what that does to the industry, but there's a, a downward trend, and this was very cleanly shown on OpenRouter with their, uh, monthly token usage, right? And when it started out, uh, most of the tokens were going through, uh, closed source models, and now about...
Zach VorhiesGUEST
It's completely inverted.
Zach VorhiesGUEST
It's like 75% of all the tokens being sold through OpenRouter at this point are being sold from open source providers.
Zach VorhiesGUEST
Well-
Mike AdamsHOST
Yeah
Marty BentHOST
the frontier model because you have the training, obviously you have the inference and obviously Anthropic, OpenAI, Meta, Google, SpaceX, they all have incredible access to this stuff.
Marty BentHOST
But I think we've discussed like OpenRouter, I think Stripe's acquisition of OpenRouter is a very strong signal.
Marty BentHOST
And I think this NeoCloud um sector is going to i i think basically having open market for different models is a very smart idea especially if we believe that being model agnostic is going to be a trend that strengthens in the future, but like to that point, like how, how do you get priority? Like, how do you both on, okay.
Marty BentHOST
Backup infrastructure side, what makes you a good cluster? What do you need to do right for your end customer to make sure people come back to you and buy compute for you? Cause there's a bunch of stuff with like networking and latency and all of that.
MarcusHOST
A new open model on Hugging Face is trending with 29 billion total parameters, but only about 4 billion active per token.
MarcusHOST
Chasing lower costs per GPU, not raw size, right? Over on OpenRouter, the top two models by token volume are DeepSeq V 4.1 Flash and GLM 5.3 Flash.
MarcusHOST
Both flash tier, both built for cheap high volume throughput, not frontier benchmarks.
MarcusHOST
Same theme everywhere.
RobinGUEST
So our mission today is to unpack exactly how this software unifies all these disparate free models.
RobinGUEST
We are looking at providers like OpenRouter, Google, Grok, and Hugging Face.
RobinGUEST
Right.
RobinGUEST
We are going to show you how to pull them all into one seamless, unbreakable API.
RobinGUEST
The connection process is pretty universal.
RobinGUEST
You go to a provider's website, generate an API key, paste it into free LLM API, and run a connection check.
speaker_3HOST
And you definitely want to start with OpenRouter, right?
RobinGUEST
Oh, 100%.
Ryan DoserHOST
Here's that Meta, that Muse Spark 1.3 model, relatively cheap in the scheme of things.
Ryan DoserHOST
But I mean, you can look at the other Metamuse models too on OpenRouter.
Ryan DoserHOST
This is just a good place.
Ryan DoserHOST
I always like to look at these types of things.
Scott RosenbergGUEST
No, download 10 different models that will fit into whatever hardware you have if you're using private AI and test them out.
Scott RosenbergGUEST
If you want, at the beginning, one of the things that I recommend for people when they're trying to figure this out, because it can take time to set up models and all of that in inference, go use something like OpenRouter or whatever, which allows you to basically use any LLM out there.
Scott RosenbergGUEST
Go to OpenRouter at the beginning.
Scott RosenbergGUEST
Go and test out 20 different models using one API key.
Scott RosenbergGUEST
Try them all out.
Scott RosenbergGUEST
Then go and take that model, the open weight model, and deploy it onto your own environment for your production labs.
Scott RosenbergGUEST
But don't go through all of the headaches in your lab environment.
Scott RosenbergGUEST
Connect that out to OpenRouter if you can, or anything similar.
TomHOST
Do you want to just explain what open router is and why you're a big advocate for it when you're building AI-driven workflows?
Phil TealeGUEST
Yeah, I like OpenRouter because essentially what it is is it's a node that points to different AI models depending on the nature of the task.
Phil TealeGUEST
So You don't necessarily need a sledgehammer for every task.
Phil TealeGUEST
Quite often a small tool will do the job just as well.
Phil TealeGUEST
You know, I want to keep the request within the UK.
Phil TealeGUEST
I don't want it going to a data center overseas.
Phil TealeGUEST
So I don't want to delegate that decision to OpenRouter.
Phil TealeGUEST
I want to have my own control over that particular step.
Type 3 AudioNARRATOR
The operation was pretty sophisticated and automated.
Type 3 AudioNARRATOR
It used three open-source AI harnesses and a tapestry of closed and open-source models via OpenRouter, GLM 5.2, DeepSeek v4 Pro, DeepSeek version 4.1 Flash, and Opus 4.6. Each attack path was chosen by the harness in real-time through extensive probing and exploitation attempts, resulting in dynamic and mostly unique attacks.
Type 3 AudioNARRATOR
It appears the hacker used minimal inputs to keep the models working, in a way similar to that used by most Vibe coders when managing their own agents.
Type 3 AudioNARRATOR
Forecasters wildly vary in their estimates as to whether a rogue AI swarm might hack a national government system of any country, or a major bank or investment firm, and demand a ransom of $10 million or more before 2028.
33 more episodes mention OpenRouter.
Create an account to see the whole feed, search across every transcript, and follow the entities you care about.