8 Comments

mettamageJun 14, 2026
https://xcancel.com/ZenMagnets/status/2065796012820848699

Correct me if I'm wrong but reading through the comments of the thread this seems to be post training/fine tuning.

KeltesethJun 14, 2026
Thanks, Firefox and uBlock does not let me watch any X content (I guess this is a good thing)
drnick1Jun 14, 2026
Same thing here, X content and trackers are blocked by my Firefox settings. The occasional inconvenience is a small price to pay not to be profiled by X, Google, FB, Amazon, and countless other Internet parasites.
oceanskyJun 14, 2026
Yes. It's post training in qwen using the novel SwiReasoning framework.
hedgehogJun 14, 2026
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.
rafaquintanilhaJun 14, 2026
Doesn't look like it's incompatible. Someone already released a quantization using MTP: https://huggingface.co/foxipanda/Rio-3.5-Open-397B-GGUF
hedgehogJun 14, 2026
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.
ramon156Jun 14, 2026
Every day I'm reminded why I don't spend time on twitter. What use does it have to claim "X is better than Y in benchmark Z, disagreeing with that means disagreeing with me"

Information is power, dick measurements are not.

reed1234Jun 14, 2026
No, I love twitter— and you are wrong.
itsthecourierJun 14, 2026
my length is a valid data point for the sake of science
adrian_bJun 14, 2026
> Post-trained from Qwen 3.5 397B

Model Card:

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

hmokiguessJun 14, 2026
Never let them know your next move
AurornisJun 14, 2026
A city government funding a fine-tune of a model is interesting.

As for the benchmarks: If you spend any time playing with fine tunes of published models you know that benchmarks are gamed so much that they're a useless indicator of performance for models from small teams. It's too easy to fine tune a model to perform well on the benchmarks, release it, put a line on your resume saying you released a model that beat the major labs on benchmarks, and then try to use that to jump into a new job. The temptation is high.

There are a lot of fringe models and fine tunes that claim to have better performance on some benchmark. Then you try to use them and find they're often worse at general tasks than the base model.

I would wait and see if these results hold across other benchmarks. It's cool that the city is doing something with AI, but this is something where extraordinary claims require extraordinary evidence. I doubt a small, previously unknown team has unlocked something secret that the team who made Qwen couldn't figure out. It's more likely it was fine tuned for a specific outcome (possibly these benchmarks) and performance in other areas was reduced as a consequence.

embedding-shapeJun 14, 2026
Indeed, this is all very true, I'd say it's true for the larger teams too, the entire ecosystem is so gamed by now that if you don't have your own private benchmarks with private test cases you haven't shared publicly, it's almost impossible to get a fair picture how well a model works, unless you actually sit down and use it.
marcosdumayJun 14, 2026
> A city government funding a fine-tune of a model is interesting.

Looks like it's an IT services government-owned company.

Most likely, they saw some business opportunity on selling it around for cities.

arjieJun 14, 2026
Benchmaxxing is the new “have a crypto trading strategy”. No one is impressed by it except non practitioners.
VoidWhispererJun 14, 2026
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'

w4yaiJun 14, 2026
> 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...

daquisuJun 14, 2026
It was a recent edit though. Yesterday snapshot: https://web.archive.org/web/20260613072958/https://huggingfa...