Who is this aimed at, I question that, it has the mixture of someone technical enough to invest in a DGX Spark and to know about Open Weight self-hostable models, but then tries to offer a sort of subscription package "we will do it for you" -- does this consumer even exist? I feel like if I owned a DGX Spark I would just build this orchestrator myself, or use some open source stuff out there, definitely would not buy into a subscription from them for this
To clear up the confusion on naming: "Computer" is their cloud-hosted, multi-model harness/router. "Portable Computer" allows you to run the harness locally. You have you bring your own DGX Spark.
It's unclear why this is tied to the Spark. Shouldn't we be able to run this on any system that is capable of running local inference?
I think we no understand why both Apple and NVidia would be interested in bringing perplexity inhouse. This hybrid architecture makes a lot of sense. Would be interesting to see how Apple responds, they are the only ones thinking about hybrid compute for long time now.
I got their Pro subscription in some promotion. Is it just me, or is Perplexity generally very bad? Specific models are often unavailable, and context collapses after only a few rounds. It frequently routes me to their Computer, which requires buying credits (upselling).
The Pro search is very fast but typically very superficial.
This got a huge eyeroll from me. Very misleading product name.
I kept skimming to find what the "device" would be but it turns out there is no device. I still have to subscribe to their services and it's just cloud infrastructure with transparent hardware. Also, it sounds like _their_ orchestrator decides what to escalate to the cloud.
Was excited for a moment for a plug and play privacy focused hardware running an agent. Would have been interesting since Apple just announced new Mac Minis today.
So: "Buy the NVIDIA DGX Spark, subscribe to Perplexity, and we'll turn it into a private local version of our Computer agent, while retaining optional access to our cloud"
Who is this aimed at, I question that, it has the mixture of someone technical enough to invest in a DGX Spark and to know about Open Weight self-hostable models, but then tries to offer a sort of subscription package "we will do it for you" -- does this consumer even exist? I feel like if I owned a DGX Spark I would just build this orchestrator myself, or use some open source stuff out there, definitely would not buy into a subscription from them for this
To clear up the confusion on naming: "Computer" is their cloud-hosted, multi-model harness/router. "Portable Computer" allows you to run the harness locally. You have you bring your own DGX Spark.
It's unclear why this is tied to the Spark. Shouldn't we be able to run this on any system that is capable of running local inference?
Nvidia Reportedly Weighs Perplexity Investment at More Than $30 Billion Valuation
https://finance.yahoo.com/technology/ai/articles/nvidia-repo...
Just let Perplexity die already.
Sour gripes from someone who can't generate $750M ARR?
Alternatively: this may be better positioned on the day Apple announced their new Minis and Studios.
This should have been the link!
https://pplx-dgx-spark-lp-demo.pplx.app/
Looks very useful
update:
It advertises with local harness and models but its closed source and you need a pro or max account
I think we no understand why both Apple and NVidia would be interested in bringing perplexity inhouse. This hybrid architecture makes a lot of sense. Would be interesting to see how Apple responds, they are the only ones thinking about hybrid compute for long time now.
I got their Pro subscription in some promotion. Is it just me, or is Perplexity generally very bad? Specific models are often unavailable, and context collapses after only a few rounds. It frequently routes me to their Computer, which requires buying credits (upselling). The Pro search is very fast but typically very superficial.
Does anyone have a good use case for Perplexity?
My definition of portable is something else :-)
Interesting that you need very expensive hardware to run this…
Sounds like a scam. No actual hardware, no reveal, just buzzwords, fluff, and fake promises.
This got a huge eyeroll from me. Very misleading product name.
I kept skimming to find what the "device" would be but it turns out there is no device. I still have to subscribe to their services and it's just cloud infrastructure with transparent hardware. Also, it sounds like _their_ orchestrator decides what to escalate to the cloud.
Was excited for a moment for a plug and play privacy focused hardware running an agent. Would have been interesting since Apple just announced new Mac Minis today.
So, what's the "luggable" version of this?
Unless I missed something, this seems to be another local agent with cloud handoff, like Warp: https://docs.warp.dev/platform/handoff/local-to-cloud/
These days I feel like Perplexity's value proposition and differentiation are slim, or not communicated well.
So: "Buy the NVIDIA DGX Spark, subscribe to Perplexity, and we'll turn it into a private local version of our Computer agent, while retaining optional access to our cloud"