Performant Models
My top performing models based off my own testing, user feedback and benches. This are the models I would recommend you using.
Text Generation • Updated • 4.36k • 147Note An updated version. Fixes the minor issues and complaints v3.1 had.
Sao10K/L3-8B-Stheno-v3.1
Text Generation • Updated • 1.85k • 101Note New and improved actual fine-tune this time, over Human & Claude-3-Opus data. I like it, more than Fimb v2.
Sao10K/Fimbulvetr-11B-v2
Text Generation • Updated • 3.1k • 127Note My current daily driver. v2 Has several major improvements over v1 in terms of non-RP tasks, and is also better at Roleplay as a side-effect. Stable context of ~10K with Rope Scaling.
Sao10K/Fimbulvetr-10.7B-v1
Text Generation • Updated • 1.5k • 36
Sao10K/Typhon-Mixtral-v1
Text Generation • Updated • 8 • 7Note A good Mixtral Merge consisting of My own models, and some top OSS ones. I like it's style.
Sao10K/WinterGoddess-1.4x-70B-L2
Text Generation • Updated • 1.39k • 19Note Practically Euryale 1.5, it is my favourite L2-70B model. An upgrade to Euryale 1.3 in terms of Non-Roleplay tasks and benchmarks, it is a more... serious model.
Sao10K/Euryale-1.3-L2-70B
Text Generation • Updated • 1.7k • 43Note Ol' Reliable. Completely Neutrally aligned model, good for Roleplay.
Sao10K/L3-70B-Euryale-v2.1
Text Generation • Updated • 1.75k • 73