Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Rio de Janeiro's city government model Rio3.5 beats Qwen3.7 in recent benchmarks

twitter.com

Post reply on HN
Print view View original

31–40 of 51 posts

Re: Rio de Janeiro's city government model Rio3.5 beats Qwen3.7 in recent benchmarks

#31

Post by w4yai »

VoidWhisperer wrote: post #30

https://github.com/nex-agi/Nex-N2/issues/4 Seems that they didn't make/train a new novel model, they did a mix of two existing models and then gave it an instruction to say it was 'Rio, trained by Rio AI Labs'

> The model is built via a merge of https://huggingface.co/nex-agi/Nex-N2-Pro and https://huggingface.co/Qwen/Qwen3.5-397B-A17B, proceeded by On-Policy Distillation from a stronger model. We detected an incorrect upload in the previous version, where the base merged version was upload instead of the final distilled model. We are sorry for the confusion and apologize profusely.

https://huggingface.co/prefeitura-rio/Rio-3.5-Open-397B/comm...

Re: Rio de Janeiro's city government model Rio3.5 beats Qwen3.7 in recent benchmarks

#32

Post by latency-guy2 »

atoav wrote: post #10
cuzezzzbbfofai wrote: post #4

[flagged]

A government ideally is a representation of the democratically chosen will of the people. If it is not, work towards making it so. IMO wherever someone says "the government" we should mentally substitute "we all, collectively". But a specific type of person appears to labour under the illusion that somehow we can get by without we all collectively steering our direction and choosing people who do what needs to be don…

"we all" is wrong, always.

You do not agree with me. You can't claim to have my interests or my will if you are against it.

Re: Rio de Janeiro's city government model Rio3.5 beats Qwen3.7 in recent benchmarks

#33

Post by hedgehog »

rafaquintanilha wrote: post #29
hedgehog wrote: post #19

Earlier quoted context omitted.

I hadn't seen SwiReasoning ( https://swireasoning.github.io , paper and code), it looks like that works at generation time without any requirements on the model. It increases token-efficiency and accuracy, but at first skim it seems like this would be incompatible with multi-token prediction. For large reductions in token budget it could be worth it.

Doesn't look like it's incompatible. Someone already released a quantization using MTP: https://huggingface.co/foxipanda/Rio-3.5-Open-397B-GGUF

As I understand it the basic premise of all the speculative decoding schemes is that the logits on the draft don't need to be exact so long as you mostly sample the same tokens, and because each position is fed by the embedding associated with the previous position's token you sort of "round away" error. With SwiReasoning I think you skip the sampling/rounding part and do something continuous using the whole distribution, so it would seem to rely on the accuracy of those values. MTP still makes sense outside the latent reasoning chunks though.

Re: Rio de Janeiro's city government model Rio3.5 beats Qwen3.7 in recent benchmarks

#34

Post by daquisu »

w4yai wrote: post #31
VoidWhisperer wrote: post #30

https://github.com/nex-agi/Nex-N2/issues/4 Seems that they didn't make/train a new novel model, they did a mix of two existing models and then gave it an instruction to say it was 'Rio, trained by Rio AI Labs'

> The model is built via a merge of https://huggingface.co/nex-agi/Nex-N2-Pro and https://huggingface.co/Qwen/Qwen3.5-397B-A17B , proceeded by On-Policy Distillation from a stronger model. We detected an incorrect upload in the previous version, where the base merged version was upload instead of the final distilled model. We are sorry for the confusion and apologize profusely. https://huggingface.co/prefeitura-rio/R…

It was a recent edit though. Yesterday snapshot: https://web.archive.org/web/20260613072958/https://huggingfa...

Re: Rio de Janeiro's city government model Rio3.5 beats Qwen3.7 in recent benchmarks

#37

Post by true_religion »

mrandish wrote: post #36

> Rio de Janeiro's city government model... Because... lack of a good open weight LLM is a pressing need high on the municipal priorities list for Rio de Janeiro citizens?

Should governments not take actions that later benefit the academic, scientific, and economic welfare of their constituents?

Or is it that it’s a city doing this?

Now Brazil does know how to boondoggle its finances for a prestigious cause with little return (e.g. the Olympics games) but this is far smaller a cost, more akin to a city setting up a tech accelerator or making a media campaign about how important STEM is.

Re: Rio de Janeiro's city government model Rio3.5 beats Qwen3.7 in recent benchmarks

#39

Post by atoav »

latency-guy2 wrote: post #32
atoav wrote: post #10

Earlier quoted context omitted.

A government ideally is a representation of the democratically chosen will of the people. If it is not, work towards making it so. IMO wherever someone says "the government" we should mentally substitute "we all, collectively". But a specific type of person appears to labour under the illusion that somehow we can get by without we all collectively steering our direction and choosing people who do what needs to be don…

"we all" is wrong, always. You do not agree with me. You can't claim to have my interests or my will if you are against it.

Yes? With sufficient pedantic spirit anything can be argued against. This is what you're doing. So to give a counter-example: You drive with three friends in a car. You ask them: "Do we all want to go to MC Donald's?"

Explain how it is wrong and why it would be. If it is always wrong it follows it has to be wrong here too. The answer is that the meaning of "we all" is context dependent and that friend of yours that argues that we all somehow includes people in the whole city is an oddball that doesn't pick up the context within the words have been said.

We can all go around and make each others day worse with deliberate pedantry by ignoring the context of words, but that is basically just a waste of human energy. If you disagree with the fundamental point I made, argue against it based on the merits of the idea instead of arguing semantics.

Re: Rio de Janeiro's city government model Rio3.5 beats Qwen3.7 in recent benchmarks

#40

Post by urbnspacecowboy »

VoidWhisperer wrote: post #30

https://github.com/nex-agi/Nex-N2/issues/4 Seems that they didn't make/train a new novel model, they did a mix of two existing models and then gave it an instruction to say it was 'Rio, trained by Rio AI Labs'

See discussion: https://news.ycombinator.com/item?id=48528371
Post reply on HN