@MrPeterLMorris which model is smaller/faster? are you suggesting everything shown there is or what? I'd be curious if qwen 27Bs are better. Don't think I'd believe that and would have to test.
also strooooooooooooongly doubt qwen3 coder next is.
Finally got Poolside's Laguna S 2.1 running NVFP4 wrangled enough to what I think is fairly benchmark it.
Model is only 72GB, which is compelling. I suspect it'll do worse than DSV4F native, but it also runs with less than half the memory requirement.
Not sure what to compare this model to.
Will also need a lot of work to get the most performance tok/sec outta this thing too, but first... let's see how smart this thing really is!
@yuvalo1212 I will certainly have it compared to dsv4f but that requires 2 rtx pro 6k. Laguna s 2.1 runs on a single pro 6k (thats also what im benchin on). We'll see! I still need to find another bench to add in for testing too.
@chris_j_paxton we tryin! it's an entirely new era with agents, and no one has any clue how to make robots actually useful either. someone will bring it to reality
@marcospereeira ok so u think 2 weeks ago when gpt 5.6 sol was topping the charts, oai wouldn't have signed? will anthropic sign once there's a moment they're not #1?
you're missing it lol
Very happy to support this on behalf of Google. We have long benefited from open source, are big contributors to open source and in fact have consistently made open weights models with Gemma available from @GoogleDeepMind@demishassabis . Onwards!
@marcospereeira I don't think so. The letter is just written to state distillation is a normal training practice and open weights are a good part of the system.
Anthropic just can't sign that based on their lobbying/statements.
They took their shot, it's failing, despite their best efforts.
@goodhunt@SarahKHeck@mkratsios47 i was gonna make a joke about how they'd just say "national security" and then I finished reading sarah's post and...what do ya know lmao.