Feels AI-generated, so much text real estate is spent explaining the basics in sufficient detail then just two paragraphs on their scheduler with a lot of hand-waving.

Given the size of the model, that's very aggressive pricing.

I wonder if it is serving at a loss to get training feedback (maybe using the margins they get from GLM 5.3 serving which is a slightly smaller model but double the price).

based on Anthropic's S1 no one is serving models at a loss, but cutting your price to keep usage high makes a lot of sense when you have your inference booked for long periods.

We still don't know that, as Anthropic don't split our costs by training vs inference.

Maybe someday.

They control the supply chain though, so they can't be price-gouged by Nvidia

they can pivot to foods: Origan, Carbonara, ....

After all they had Nano Banana for years.

AI would discover and name new elements by that time. Getting humans out of the loop of the naming problem.

Not sure humanity would be around after 118 models...

> For ultrafast latency, only activations need to be passed between the wafers, and this can be done via RDMA over Ethernet.

I believe it for decode but the article says nothing about how they make prefill fast on that interconnect story :/.

Still 9 days before the release of weights apparently.

Amazing news, I remember the days when you had to fight XLA to make PyTorch run on TPUs, it seems we finally got first-party support!

Since HW specifics live mostly in the kernel layer (Pallas or CUDA), I imagine the Jax vs PyTorch question is more of a style preference now?

Schools got better at teaching programming but smartphones replaced the PC culture.

I don't think I'd be currently doing software if YouTube and games were accessible and good on phones back then.

1.5 TB/s 1.3 PFLOPS with no fast multi-GPU, it's a consumer card that pretends to be a server card.

It would work great on Qwen3.8-Flash-Next or GLM-5.3 as single user decode only hits a few experts, anything with concurrent users and you'd be sitting for a while.

That's the point I believe fast local 2-4 user service for cheap on very large models.

You have to pretend it's for servers to keep investors happy, it would be banks and similar with in house servers that have to run local large models buying these for experiments and cybersecurity. Likely.

Reminds me of the early 2000s era where we'd fight in my household if too many people used the internet at the same time, can see that happening here if it's not 2 to 4 users but 10.

I mean in companies you can just buy more, and I think AI compute is going to get a lot cheaper in the next 4-5 years, photonics are just around the corner, assuming AI is even 1/10th as good as AGI they should help prove substantially useful for just pushing photonics based computing over a couple decades sooner.

Not really, well depends on the price point but it would make more sense given enough users to multi-GPU setup with good interconnect between them to do expert parallelism.

Fun to see the LLMism on the product page: "A, not B"

Productivity!