Interesting new model: Motif-2-12.7B-Reasoning

Posted by LoveMind_AI@reddit | LocalLLaMA | View on Reddit | 5 comments

I didn’t see much discussion of the instruct version, but the reasoning version is out and it sounds like an interesting model. They were not on my radar until recently. Any thoughts? I do think models in this size range seem to look more and more like the future. https://huggingface.co/Motif-Technologies/Motif-2-12.7B-Reasoning

Reply to Post

5 Comments

[-]

LoveMind_AI@reddit (OP)

OK, I've spent more time with it this AM. I'm honestly impressed. Very naturalistic for english conversation - extremely coherent and able to hang with complex concepts (like a simulated identity configuration prompt embedded in convoluted pseudocode) that threw Hermes 4.3 36B and Olmo 3.1 for a major loop. It's a cool model! I like it a lot better than Apriel-1.5.

[-]

LeatherRub7248@reddit

is there a cloud api provider for this? their direct chat site seems quite limited

[-]

LoveMind_AI@reddit (OP)

I haven't found it yet. Yeah, there's no point in actually chatting with the thing longterm on their website. But I was able to do enough of my baseline tests to get a sense of how it reacted and I'm impressed that it doesn't feel like every other single model out there.

[-]

Odd-Ordinary-5922@reddit

no point even using the reasoning model when the instruct is that close

[-]

Cool-Chemical-5629@reddit

Model's chain of thought feels endless. I tested Motif some time ago (they have an official chat website where you can chat with the model, so that's a nice bonus), but the model itself did not really impress me. Maybe it's the smaller size, so it may not be suitable for your specific use cases and I know it failed when I tried it with mine. I believe it's a Korean model, so hopefully Korean users will find it useful and perhaps better for their native language than other models in that size range.