Rendered at 03:05:19 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
itkovian_ 5 minutes ago [-]
The most important thing here is training was entirely nvfp4 - this is best model by far that has ever been trained in that precision and explains why they were able to do it so fast
Jackobrien 2 hours ago [-]
The best open models today have been “good enough” to do most knowledge work for a few months. This might be the first US one good enough too. That’s a big deal.
Open weight US based Opus 5.5 within 12 months.
josh-wrale 14 minutes ago [-]
Muse Glimmer is pretty good.
npc99 1 hours ago [-]
Inception Mercury 3 and Particle-1's model are also good.
(I am assuming Inception is an American company, not sure TBH)
ray__ 45 minutes ago [-]
Agree, Mercury 3 is one of my favorite models right now...the speed is such a quality of life improvement.
martincsweiss 27 minutes ago [-]
Really? I'm surprised. What do you use it for?
My read on Diffusion LLMs has been that it's cool tech, but only useful for things like like code autocomplete or editing a block of text.
They don't beat autoregressive models on quality per compute, assuming you are serving lots of users and can batch requests.
serf 28 seconds ago [-]
MoA schemes that use proposers/task-separation/supervisory roles can still benefit from very fast diffusion models like mercury, the overall speed doesn't have to be bottle-necked by the slowest model -- that depends on the arrangement it's deployed in.
but to your point , yeah diffusion models can be kind of brain dead without proper coaxing.
kadoban 2 hours ago [-]
> Open weight US based Opus 5.5 within 12 months.
Seems unlikely. Maybe open weight at all near there? That'd be great.
wolttam 1 hours ago [-]
People say this about the frontier every time, and the open models catch up in a matter of months every time.
kadoban 58 minutes ago [-]
The open models catch up, the _US_ open models catch up much more slowly. Different incentives and cultures.
p-e-w 2 hours ago [-]
Why does that seem unlikely? The difference between frontier and open weights has been much less than 12 months for a while now.
satvikpendem 2 hours ago [-]
US open weights are much worse, way more than 12 months apart compared to frontier.
kadoban 1 hours ago [-]
> Why does that seem unlikely? The difference between frontier and open weights has been much less than 12 months for a while now.
Frontier and open weights? Yes.
Frontier and _US_ open weights? No.
I'd expect a US open weights provider to suddenly get bought or go non-open if they get near Opus 5.5 anyway.
nghiemphucminh7 45 minutes ago [-]
[flagged]
epsteingpt 2 hours ago [-]
It's a nice, domestic open weight model.
Without hundreds of millions in 9 days a small team is producing a model that can beat DeepSeek 4.1 and a platform to tune it.
Optimistic for the future.
a11r 2 hours ago [-]
Very refreshing to see benchmarks where the reporting team is not the clear winner. They are claiming to be better than GLM 5.3 flash, which is impressive if it holds up outside of benchmarks.
tcdent 3 hours ago [-]
Is it SOTA yet? No. Are the domestic open weights labs demonstrating incredible velocity? Undeniably.
bashtoni 3 hours ago [-]
Distillation is the tide that lifts all boats
epsteingpt 2 hours ago [-]
Imagine if Mira Murati and John Schulman actually know what they're doing?
tcdent 2 hours ago [-]
Given the time to market and the date in which these labs were founded, I find it far more likely that the lag was representative of them establishing proper internal (RL etc) tooling.
Neywiny 1 hours ago [-]
I feel like this is a dumb question, but why are we spending any effort on wine comparisons? Is there a way to get rid of experts on these to just be ones for code?
randomblock1 28 minutes ago [-]
It's not like they're training on that. That's just a good showcase of how the model is good at generalizing & understanding (not benchmaxxed). You might want something similar to these experts for tasks like detecting anomalies or discoloration or whatever.
Neywiny 16 minutes ago [-]
Source they didn't train on it?
kccqzy 1 hours ago [-]
What if your coding project is related to wine? So you would want facts about wine to be inside the weights, not through web search.
Neywiny 43 minutes ago [-]
If it was about wine, why would I want one with a training data cutoff? I'd much rather have one that can query a database about wine
syntaxing 2 hours ago [-]
I can’t seem to find the cost? The craziness of deepseek flash 1017 is how ridiculously cheap it is for how capable it is.
Bolwin 2 hours ago [-]
AA lists it as $0.30/0.03/1 and cost per task is $1.18, about double that of deepseek
syntaxing 2 hours ago [-]
It’s about 2-3X more expensive so about the same token efficiency as 1017. I can see this being a good option for the US.
kadoban 2 hours ago [-]
Wow, that's pretty good.
ddxv 2 hours ago [-]
This is amazing! Can't wait to trying this out. Also, crazy that I found this on LinkedIn with more upvotes and discussion than the HN thread has.
lostmsu 2 hours ago [-]
> Small: 594 billion parameters
cries in GPU racks
colingauvin 2 hours ago [-]
You could probably run a Q2-Q3 with 2x DGX Sparks. So...$12k at today's prices. Plenty of folks are running DS 4.1 Flash with Strata and some GPU + bulk in system memory. Maybe more like $3k-$4k? But speeds come down quite a bit.
jazzyjackson 2 hours ago [-]
I don’t like it. Poor form to squat on an old companies name like that, one which has a rich legacy in AI. I know the trademark was cancelled so all I can say is I don’t like it.
singpolyma3 2 hours ago [-]
So basically comparable to DS4.1? But why bother to make something we already have?
multisport 1 hours ago [-]
Because its american.
singpolyma3 1 hours ago [-]
lol
reasonableklout 60 minutes ago [-]
Sovereign open source AI is a real thing, lots of companies want to own their own intelligence but also don't trust the models coming out of China.
singpolyma3 21 minutes ago [-]
Why would you trust anything coming out of USA?
estearum 15 minutes ago [-]
The vast majority of users on this website wouldn't be trusting something coming out of the USA, they'd be trusting something in the USA.
Open weight US based Opus 5.5 within 12 months.
(I am assuming Inception is an American company, not sure TBH)
My read on Diffusion LLMs has been that it's cool tech, but only useful for things like like code autocomplete or editing a block of text.
They don't beat autoregressive models on quality per compute, assuming you are serving lots of users and can batch requests.
but to your point , yeah diffusion models can be kind of brain dead without proper coaxing.
Seems unlikely. Maybe open weight at all near there? That'd be great.
Frontier and open weights? Yes.
Frontier and _US_ open weights? No.
I'd expect a US open weights provider to suddenly get bought or go non-open if they get near Opus 5.5 anyway.
Without hundreds of millions in 9 days a small team is producing a model that can beat DeepSeek 4.1 and a platform to tune it.
Optimistic for the future.
cries in GPU racks