JRizzAi_Labs

Local LLM deployer | Privacy-first AI stacks | Sharing real deployments, opinions & lessons | Here to learn and help others run AI locally

[JRizzAi_Labs @jonathantrizzo]

@TechMDAI By December the price will go up to $620 from $520 and then week before Christmas there will be a $100 holiday discount.

[JRizzAi_Labs @jonathantrizzo]

GB300 Workstation and then I would slowly add AMD Halos nodes per qwen3.8-flash-next distilled from glm-5.2 runs on the GB300 if not something else I prefer.

Create a fleet of specialists and a teacher model constantly working on the gb300.

Although I may be super satisfied with glm-5.3-flash nvfp4 on max thinking in a lot of ways

[JRizzAi_Labs @jonathantrizzo]

@ivanfioravanti So i was LM Studio for the past year. I've played with oMLX and MTPLX and since I haven't moved from qwen3.6-35b-a3b I haven't felt the need to swap it up.

Who has feedback on vLLM-MLX?

[JRizzAi_Labs @jonathantrizzo]

For now

[Frank Azor @AzorFrank]

BTW @AMDRyzen AI Halo with 128GB is still available for $3999: https://t.co/eArMoz6buz

[JRizzAi_Labs @jonathantrizzo]

@digitalix Almost like that was the intent

[JRizzAi_Labs @jonathantrizzo]

@TechMDAI If local AI enthusiasts are spending $12,500 for 256gb M5 Ultra I can see why they increased the price to be closer to them.

[JRizzAi_Labs @jonathantrizzo]

@TechMDAI It’s a solid deal for sure and wish I could secure it but I need a couple of things to fall in place.

I’ve seen quotes for the Dell GB300 at $180k. Don’t get that.

[JRizzAi_Labs @jonathantrizzo]

@Priyannkaaaa I’ve been using https://t.co/gmanBrVRn0 coding plan for over a year and have been very happy with their work. I would like to try all of these models but I would prefer to support https://t.co/gmanBrVRn0

[JRizzAi_Labs @jonathantrizzo]

Very surprised to see these results and I hate to ask, but I wonder if there’s a shift with 2 x M5 Uktras vs 4 x DGX Sparks. @nvidia is clearly in the lead still.

[Alex Ziskind @digitalix]

Dual DGX Spark cluster vs M5 Ultra Mac Studio - GO! (new video is out)

[Alex Ziskind @digitalix]

Dual DGX Spark cluster vs M5 Ultra Mac Studio - GO! (new video is out)

[JRizzAi_Labs @jonathantrizzo]

@alexocheema @Apple @Apple 👏

Now send me ☝️

[JRizzAi_Labs @jonathantrizzo]

My M3 Ultra 256GB - This is development that makes me excited for NOT selling in the aftermarket. Clustering over the years, if scalable will be great for local AI deployment and development.

[0xSero @0xSero]

We've made an inference breakthrough at Exo.

You'll hear more Late October.

[JRizzAi_Labs @jonathantrizzo]

@0xSero M5 Ultra Scaling bandwidth???

DGX + M5 Ultra scaling bandwidth?

[JRizzAi_Labs @jonathantrizzo]

I have finally decided to rent 8 c RTX pro 6000s to distill glm-5.2 in to some fine tuning training data for qwen3.5-122b-a10b and some smaller models.

So that’s next and following should clear the path for launch. All profits from what I drop will be going towards a Supermicro GB300 workstations and smaller ai inference nodes. (GB10s, older Mac Ultras)

[JRizzAi_Labs @jonathantrizzo]

@TechMDAI I dream of nvfp4 on a gb300 workstation. That’s my “I’m good right here” dream

[JRizzAi_Labs @jonathantrizzo]

@_Gibsonmel This is AI all

[JRizzAi_Labs @jonathantrizzo]

@sudoingX Not in the form of caring about inference or downloading models.

Just a packaged device that can be reached from mobile phones

[JRizzAi_Labs @jonathantrizzo]

https://t.co/bAWz4uflPo