Super-Slop-Machina-Roleplay-1.2b-V4-ERP-Tolerant-it

Super-Slop-Machina-Roleplay-1.2b-V4-ERP-Tolerant-it: Post Trained Release of The Super-Slop-Machina-Roleplay-1.2b-V4 After SFT

"Its finally ready."

Improvement:

• Adapted for longer context
• Consumed higher quality data

And what does that mean?

Way, way better roleplay than the preview version.

Quants(Speaking from personal experience with this specific model):

• BF16- Recommended, highest quality, least logical mistakes.
• Q8_0- Recommended, high quality, makes slightly more mistakes but nonetheless near lossless.
• Q6_K- Recommended if Q8_0 is too much, degradation begins, not exactly notable here, but you will notice minor detail loss.(When Mradermacher quantizes this model, I recommend getting his i1 Q6_K quant instead of the one I got in my repo, but in any case I still recommend Q8_0 or BF16)
• Q5_K_M- Recommended if hardware is really, REALLY bad, degradation becomes noticeable.
• Q4_K_M- Not recommended for most use cases, degradation is clearly noticeable.

Quants can be found in the repository, along with safetensors.

Note:

I finally somewhat succeeded with gathering my best chatlogs. I don't yet know whether it turned out better than the last version but I do believe its quite better now.

Under 1B models soon! I tried my best with them, results were rather unusual, when I'm done with them, I'll probably return my attention back to ministral 3 3b.

Big announcement:

I experimented with Qwen 3.5 9b and I think I am somewhat succeeding, I will soon release the finetuned weights, result that I was able to achieve with it was VERY interesting...