Any ETA on the bigger (31B, 26B) Gemma 4 models?
Hey, just recently discovered your work and absolutely love it! Was just curious when you expect to upload the bigger Gemma 4 models (31B, 26B) :)
Awesome job!
Nvm, just saw your comment in a similar discussion!
To anyone else wondering, HauhauCS said the bigger models should be ready tomorrow today! :D
Nvm, just saw your comment in a similar discussion!
To anyone else wondering, HauhauCS said the bigger models should be ready
tomorrowtoday! :D
Hello - reopening this to offer an update (and so that others can see it as well) - E26B-A4B is done (if all goes well) within the next 9 hours (watching it as we speak). After that the dense model in another day or two :)
So, when we wake up tomorrow morning there will be surprise ready for us? cool!
Please create discord server for announcements and discussions. Thanks
Nvm, just saw your comment in a similar discussion!
To anyone else wondering, HauhauCS said the bigger models should be ready
tomorrowtoday! :DHello - reopening this to offer an update (and so that others can see it as well) - E26B-A4B is done (if all goes well) within the next 9 hours (watching it as we speak). After that the dense model in another day or two :)
I won't be awake for 3 hours longer sadly, will upload it today (Saturday) :)
I did not like all the outputs so far so I want to make sure that the releases maintain the quality everyone has gotten used to.
I'm also in favor of you setting up a Discord server for announcements, testing, and discussion. With so many followers, that would be really interesting, Of course, I get it if your schedule is too packed to manage it.
Thank you for all your hard work.
Hello, as an update - still working on them :-)
For a Discord server, I'll give it a good think. It'll be nice to be able to keep everyone updated so I might make one as soon as I clear up a lil' bit. Got a massive backlog now (the 2 Gemma4 models, upcoming Qwen3.6, upcoming Minimax M2.7 which I hope I'll fit in my VRAM, and my new project which I really really think it'll change the game for local models in specific use-cases, if I manage to nail the execution)
hello. can you use Unsloth's scripts to make a UD quant for the larger models? I would love to see a UD-IQ4_NL quantization if possible (it fits well on 16gb VRAM and has decent quality). Thank you.
Hey, thanks for the update. Just checking in, do you have a rough ETA for the E26B-A4B release? Even a broad estimate like today / tomorrow / this week would help a lot.
Hey, thanks for the update. Just checking in, do you have a rough ETA for the E26B-A4B release? Even a broad estimate like today / tomorrow / this week would help a lot.
Im also super excited for it. He mentioned somewhere he wasnt satisfied with some answers and continues working on it though.