35
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 26 Aug 2026
35 points (100.0% liked)
LocalLLaMA
5087 readers
56 users here now
Welcome to LocalLLaMA! Here we discuss running and developing machine learning models at home. Lets explore cutting edge open source neural network technology together.
Get support from the community! Ask questions, share prompts, discuss benchmarks, get hyped at the latest and greatest model releases! Enjoy talking about our awesome hobby.
As ambassadors of the self-hosting machine learning community, we strive to support each other and share our enthusiasm in a positive constructive way.
Rules:
Rule 1 - No harassment or personal character attacks of community members. I.E no namecalling, no generalizing entire groups of people that make up our community, no baseless personal insults.
Rule 2 - No comparing artificial intelligence/machine learning models to cryptocurrency. I.E no comparing the usefulness of models to that of NFTs, no comparing the resource usage required to train a model is anything close to maintaining a blockchain/ mining for crypto, no implying its just a fad/bubble that will leave people with nothing of value when it burst.
Rule 3 - No comparing artificial intelligence/machine learning to simple text prediction algorithms. I.E statements such as "llms are basically just simple text predictions like what your phone keyboard autocorrect uses, and they're still using the same algorithms since <over 10 years ago>.
Rule 4 - No implying that models are devoid of purpose or potential for enriching peoples lives.
founded 3 years ago
MODERATORS
So this was the Ox-Alpha model on OpenRouter. That’s cool.
I guess the sleuths were right. I've used local GLM models in the past, they aren't bad at all. although I'm currently enjoying the speed and quality of Ornith 1.5 35b running on HaloFPX.
just wish there was an abliterated version.
There are several abliterated versions listed in the finetunes: https://huggingface.co/models?other=base_model%3Afinetune%3Aornith-ai%2FOrnith-1.5-35B-A3B
I haven't tested any of them personally though.
oh I know, but since I'm running on a Z13, there's a custom runtime built for Strix Halo that makes the model a bit faster. and that runtime uses a fine-tuned version of Ornith that takes advantages of the specific architecture to increase t/s. I know, it's already fast, but it's nice to see those first tokens start generating within a couple seconds, and my agent crank out a few thousand tokens in less than ten seconds.
ornith isn't good for coding, sure, but for automating the boring shit I don't want to pay attention to, the model works beyond expectations. I do wish it was abliterated because one of the things I use it for is OSINT workflows and tracking down the personal info of fash, and the guardrails version doesn't always like to do that, so the skills have to carefully word the prompts. I imagine it would get pissy if I asked it hard questions about China too, but I don't chat with LLMs so I'm not sure.