Note Lol. I should do a 4B lora with this and do some shitposting

Note Lots of Claudeisms. need to extract, rewrite.

Note Same thing as above. Lots of ISMs. Like cocksleeve.

ICTNLP/Llama-3.1-8B-Omni

9B • Updated Nov 14, 2024 • 28 • 418

Note Need to release how they did this.

nvidia/Llama-3.1-Minitron-4B-Width-Base

Text Generation • 5B • Updated Feb 14, 2025 • 5.23k • 197

Note EXTREMELY good for the size, Better then Llama3 8B for Roleplay/Creative writing and at half the param size

Note Could be cool if you synthmaxx with certain interesting personas.

Note Could regenerate with something better then GPT3.5 - would love a model to be better at french.

Note all three look interesting, need second opinion on whether slop or not.

FourOhFour/Vapor_v2_7B

Text Generation • 8B • Updated Apr 27, 2025 • 6 • 5

Note Cool finetune of Qwen2.5 7B makes me reconsider what to do again. Instruct was coal but this seems nice? base worth saving?

Note all of these look really nice, but each train i've used them - they've ended up causing the model to only be coherent at temp2, minp0.3 - need lower LR and recook with this

Note Remove open hermes samples.

allenai/Llama-3.1-Tulu-3-8B

Text Generation • 8B • Updated Feb 13, 2025 • 5.89k • • 178

Note L3.1 was coal - could this be a better base to finetune upon? don't like the DPO step...

nvidia/Hymba-1.5B-Instruct

Text Generation • 2B • Updated Jan 2, 2025 • 359 • 245

Note Looks interesting, too bad GGUF will never support it

Djrango/Qwen2vl-Flux

Text-to-Image • Updated Dec 6, 2024 • 508

Note Quants? looks neat

Qwen/QwQ-32B-Preview

Text Generation • 33B • Updated Jan 12, 2025 • 34.5k • • 1.74k

Note COT is a meme, this will go nowhere but might be neat if it was a smaller size

ATH-MaaS/Marco-o1

Text Generation • 8B • Updated Nov 23, 2024 • 618 • • 713

Note Chain of Sex thoughts?