coder543

There is also a new, larger voice to text model available... but not for an old iPhone like the 13 mini. Only like the iPhone 17 Pro and above.

The best workaround is a third party app that lets you run Parakeet V3 or Whisper on your phone. There are quite a few.

Please run GLM-5.3 and GLM-5.3-Flash. I would love to see how they do. On the smaller end of things, Qwen3.8-27B and Ling-3.0-Flash would also be interesting.

In the benchmark, have you considered instructing the models to build their own SPICE simulations to test their work? Simply asking them to write and run simulations could improve performance, even without telling them what to simulate.

Novita does not offer Hy4-preview on either OpenRouter or their own model list. Maybe you confused it with Hy3.

The spark can easily run UD-Q4_K_XL on this model... using IQ1_S doesn't make much sense.

No. What else could it reasonably be named? Hard to imagine.

The rule has always been intended to cover types that have another word in them but still choose to pointlessly repeat the package name.

`uuid.UUIDGenerator` is a hypothetical example of the anti-pattern that would instead be better named as `uuid.Generator`.

I think glancing at a random snapshot from today misses all the context. Nemotron 3 is far more significant than you're giving it credit for.

At this point, Nemotron 3 is really an 8 month old model series. That's when Nemotron 3 Nano was released, and the Nemotron 3 Super/Ultra models this year are obviously based on that recipe, mostly just bigger with a few tweaks here and there. Against today's models, no, not that interesting. Each of the Nemotron 3 models were briefly competitive when they launched, but never exceptional, and less competitive with each scale up. The fact that it took so long for Nemotron 3 Ultra to launch really hampered its competitiveness.

The Nemotron 3 series is extremely open about training recipes and training data, far more open than most open weight models, and that is valuable.

Before Nemotron 3, Nvidia had never released a single LLM that I would consider interesting at all, so Nemotron 3 was a big step up. The closest thing was Mistral NeMo, but a significant part of the credit there goes to the Mistral team, not Nvidia.

Given how much Nemotron 3 improved, I'm curious to see if Nemotron 4 will take them to a leading edge level instead of just briefly competitive.

(Nvidia released a Nemotron 3 and a Nemotron 4 like 3 years ago... this year's Nemotron 3 is entirely unrelated. Nvidia's naming schemes leave a little bit to be desired.)

Yes, the article uses AI-assisted text, so if that offends you, no need to accuse: you can just stop reading.

I have a good amount of background knowledge on all of this, and I've invested hours into pulling this short series together, so this was not just some single shot prompt and hope for the best.

As someone who has used LLMs since ChatGPT first launched, and used them locally since the first open models were released around the Llama 2 era, I have a perspective to share that I think offers some value. LLMs aren't magic. People who have only ever used full reasoning models with agentic tool calling loops may not appreciate how they work under the hood, so this article series tries to give you the foundational concepts and terminology to do more research. Hopefully, it also gives a basic understanding.

Part of the difficulty of putting something like this together is that I don’t know what people don’t know, so there will invariably be gaps that make some of this hard to understand.

I am sure there are ways this series can be improved, so any constructive feedback is welcome.