Where is MiniMax M2.7?
Posted by lolwutdo@reddit | LocalLLaMA | View on Reddit | 13 comments
Feel like it’s been forever since they announced it; I wonder if it’s going to be better than Qwen 3.5 397b.
Posted by lolwutdo@reddit | LocalLLaMA | View on Reddit | 13 comments
Feel like it’s been forever since they announced it; I wonder if it’s going to be better than Qwen 3.5 397b.
xw1y@reddit
This weekend expected
lolwutdo@reddit (OP)
It’s the weekend already, unless they plan to do a release on Sunday which is a weird day
xw1y@reddit
It’s releasing today
DeepOrangeSky@reddit
Haven't there been some other major Chinese models that released on a Sunday night in American time zones (Monday in China)?
I remember some people mentioning it when the Qwen3.5 models were looming and people were arguing with each other on here about whether a Sunday release day was plausible or not. I can't remember what the major Chinese model/models was/were that dropped on a Sunday that they made reference to, like if it was DeepSeek or an earlier Qwen model, but I think it was a major one that had dropped on what was a Sunday night in American time zones or something.
Zyj@reddit
Did you miss this post? https://www.reddit.com/r/LocalLLaMA/comments/1se8175/minimax_27_good_news/ pointing to this comment https://huggingface.co/MiniMaxAI/MiniMax-M2.5/discussions/53#69d3e884ba6f6793d723f30e ?
lolwutdo@reddit (OP)
Nope, it’s the weekend and no word hence a the post.
suicidaleggroll@reddit
Their original post said 2 weeks, 2 weeks was last Sunday (6 days ago). That day or the day after, there was a new post saying they’re still working on it but there was a bit more to it than they had originally hoped, so it would take a bit longer. This week or next most likely.
LegacyRemaster@reddit
Fingers crossed between now and tomorrow. It looks like Qwen is abandoning open source.
eclipsegum@reddit
Qwen 3.5 397b is already so freaking good. It’s my daily driver and my Claude usage is down 90%
RipperFox@reddit
Did you try GLM 5.1? It uses tokens very efficiently, I'm very impressed..
redblood252@reddit
That good? What do you use it with for agentic coding? How good is it compared to opus at analyzing code bases for refactor and fixing bugs?
ScoreUnique@reddit
Yeah would like to get opinions about how does it compare to opus :)
I have 192gb ram and 48gb VRAM so I can run the model at 5-7 tokens gen but for agentic stuff it's painful.
Geximus-therealone@reddit
Hope MiniMax is fulfilling its promises, yet to wait for quality quantinization