maldv/Loqwqtus2.5-32B-Instruct

Text Generation • 33B • Updated Mar 19, 2025 • 3 • • 2

Note Qwen3.2 -> fine-tune EVA -> fine-tune CoT model. Tested quickly, seems okay. ChatML + Think tags

google/gemma-3-27b-it

Image-Text-to-Text • 27B • Updated Mar 21, 2025 • 1.65M • • 1.87k

Note Meh. Extremely memory hungry at 16K context, unless you quant the KV cache which comes with its host of issues. It's to a point it's not worth testing for me (and neither are any finetunes by extension)

LatitudeGames/Harbinger-24B

24B • Updated May 16, 2025 • 74 • 79

Note MS24B 3.1 (Base) fine-tune. ChatML. Trained on narrative at the second person *you do this*. Sounds good.