On 1 and 3, the obvious move is to shift the bulk of the harness behind a new API that's not based on raw LLM access. Then they get to hide secret sauce behind that API and all three go from commodity to premium while simultaneously being able to try out whatever tricks they can get away with to reduce their own inference costs. I'm almost surprised this hasn't happened already.
Bedrock is really out of date with the models it offers, to the extent that I'm not sure they even have plans to update what's on there now they have the deal with Anthropic. They're still offering Qwen 3, not even 3.5 and certainly not 3.6. GLM 5 is the newest z.AI model they have, when it's 5.2 that would be the one to worry Sonnet.
There are some ok models on there (Qwen 3 Coder Next is usable and fast, for instance) but the lack of updates in a fast-moving field makes it something I don't want to recommend to my org.
1) the company has device-level control to the degree that they can not only restrict which API endpoints people can connect to but which accounts they use to do so (in which case this already isn't an issue); or
2) they don't, and all bets are off anyway, open weights or not.
With this model size I've found that the harness seems to matter more. I've moved on to little-coder rather than raw pi with qwen3.6 27b personally, it might be worth taking a look.
Nah. There are already models at every size on the scale. If you want to run an open 1T model today, you can.
What's going to happen is that the capability at any given size point is going to get better over time as new training regimes cram more into the available space. A 27b model released next year will be better than a 27b model this year (else why release it?). Hardware will get more useful, not less.
That ends up begging the question, because the next step is "how high do you have to drop it from so that it's travelling twice as fast?" and you're immediately going round in circles.
"Eventually" doing a lot of work. Micron (and implicitly anyone signing this deal) are betting demand is going to outstrip capacity for several years, taking into account what new capacity can be brought online and when.
I do think this is worth emphasising: the article only focuses on mortality. Not quality of life. Vitamin D makes me not feel like crap, it's cheap, and effectively zero risk. I'm not expecting it to make me live longer, I like that it makes me live better.
You could, but it's driving in the wrong direction to try to build that knowledge into the model weights because you'll always run into a capacity limit sooner with a small model than with a larger one. The thing the model is specialised for is linguistic understanding and the reasoning process itself, and you max that out at the expense of domain-specific knowledge. If you take "as few weights as possible" as a given, I think the interesting question is how small you can make the model with externalised memory. The openclaw and hermes people are all over this sort of memory problem: using the local filesystem or a local database of some sort is exactly a "very fast local memory" where the more you use it, the more knowledge it gathers. Whether that translates to it being "smarter" is a deeper question than it looks.
That particular example doesn't quite fit, but I've certainly seen cases where otherwise perfectly ordinary fixed strings needed to be broken up to meet linting rules.
"Pretty much" doing a lot of work. But it's kinda analogous to 99% JPEG compression: yes you can detect loss, but you get meaningful compression ratios out of it and the subjective appearance is nigh-on perfect.
Shannon would be pointing out that if you can throw away half the model without apparent degradation, we're nowhere near packing in all the information we could in training. There must be a better arrangement than we've currently got.
The problem is that the situation in the RAM market might just... not go away. It's locked in for the next couple of years unless the AI market goes pop. Which it might! But if it doesn't, there's no particular reason to think that the incentives for cornering the market like OpenAI have would go away.
We might see that new normal in five years or so. We will see a new normal sooner than that if there's a run on AI because of the sudden availability of DRR fab capacity, but also we'll probably see the level of local models freeze at whatever state they've got to at that point. But an equally likely outcome is that any new DDR capacity that comes online is just immediately absorbed by frontier AI, and consumer devices stay at "just good enough" for a decade.
I've been getting 40-50t/s out of qwen3.6:27b on a 4090 limited to 350W with the MTP changes that went in. That comes out at 8.75J/t at the upper end. No idea how that compares with anything else out there. I'd expect a 5090 to be a bit cheaper because it'd be faster within the same power limit.
The difference won't be in the individual tasks. It'll be in the scale of job they can take on and how you interact with the model. Think of pairing with a junior vs replacing a full delivery team, that's the sort of difference we'll be looking at. We'll be able to get closer to the latter by being more clever with harnesses, I reckon, but the frontier labs will run ahead because for any given harness trick they can lean harder on model smarts.