Just posted another blog post on @huggingface, this time about new tensor-type layouts in llama.cpp 🤗 huggingface.co/blog/bartowsk… I shout them out in there but I'll do it again here, this research was largely powered by @FrameworkPuter Desktop 👀

Made a short blog post about most recent imatrix dataset experiments :) Enjoy! huggingface.co/blog/bartowsk…

Anyone know what happened to Yorkie..? I guess I'm a bit late to the party, but seems they went and deleted their discord, X, and github :( if you're out there Yorkie, hope you're okay, feel free to DM or email, just to let me know you're alive

Introducing the Trinity Builders Program. This is a community credit grant that gives developers, researchers, and open-source builders free inference access to Trinity models on our API. Why we're doing it: Open weights only go so far if you don't have the compute to actually build with them. We want to lower that barrier so you can run experiments, iterate, and ship real systems. Who this is for: - Open-source projects - Research initiatives - Prototypes and production applications - Developer tooling Grants range from < 50M tokens to > 1B tokens, allocated based on your submitted project. Submitting an application does not guarantee approval. We'll review requests on a best-effort basis, and allocation depends on compute capacity, token budgets, and other operational priorities. We can't wait to see what the community builds with Trinity. Learn more and apply here: arcee.ai/blog/introducing-th…

Today we're releasing Trinity-Large-Thinking. Available now on the Arcee API, with open weights on Hugging Face under Apache 2.0. We built it for developers and enterprises that want models they can inspect, post-train, host, distill, and own.

Forgot to update the Qwen 0.8B - 9B models with the F32 ssm_alpha and ssm_beta weights, redoing now :) If you were getting lower performance, in particular speed-wise, it may do you well to redownload (especially on strix halo apparently?)

Alright, I've added the mean and 99.9% KLD, and the mean and same top P statistics to huggingface.co/bartowski/Qwe… README This is more of a one-off, but I will explore doing it more regularly in the future, just gotta figure out how best to handle with my limited GPU compute :)

One (hopefully) final update to the Qwen3.5 model quants.. If you're enjoying the model, don't bother redownloading :) But basically with inspiration from Aes Sedai and ddh0 and others, I just tweaked the recipe a bit to be much simpler and it seems more stable, will post KLDs

Thinking of another small switch up for my model naming I want to make a list of "official" model releasers, Qwen, Deepseek, Arcee, Google etc, and when they release a model, I don't include the Author_ naming Everyone else continues with the Author_ for clarity Thoughts?

A banner weekend for Arcee as Trinity Large Preview officially entered the Top 5 overall models on @OpenRouter! After surpassing 80B tokens on Sunday, we are already on pace to break that record again today. As our inference volume consistently tests the upper bound of our current limits, this growth signals a compounding market demand for our frontier-class open weights as we scale far beyond our initial 1T token milestone. Stay tuned—more exciting news is on the horizon!

The usage from Trinity, especially while free on openrouter, has been incredible. We’ve actually managed to handle 3x more throughput on our inference cluster than we estimated. But our TPS has tanked as usage just continues to grow. We moved some training compute over to inference (not as simple as I thought it’d be) and we should be back to full speed. Give us all you got.