zai-org/GLM-4.5

Text Generation • 358B • Updated 22 days ago • 95.3k • • 1.28k

Note My recommendation for the big-several-hundred-B MoE size class. I run it in non-thinking mode with assistant prefill. It's a bit too slow to run in thinking mode locally on CPU+GPU.

Doctor-Shotgun/ML2-123B-Magnum-Diamond

Text Generation • 123B • Updated Jul 7 • 16 • 8

Note Focusing on applying enough heat and pressure to dry, assistant-tuned models until they turn into creative writing gems!

Doctor-Shotgun/L3.3-70B-Magnum-Diamond

Text Generation • 71B • Updated Jun 25 • 221 • 3

Note Focusing on applying enough heat and pressure to dry, assistant-tuned models until they turn into creative writing gems!

Doctor-Shotgun/MS3.2-24B-Magnum-Diamond

Text Generation • 24B • Updated Jul 7 • 2.33k • 35

Note Focusing on applying enough heat and pressure to dry, assistant-tuned models until they turn into creative writing gems!

Doctor-Shotgun/L3.3-70B-Magnum-Nexus

Text Generation • 71B • Updated May 12 • 1.24k • 8

Note This is a merge of my various L3.3 Magnum tunes, leading to a more stable result. Works with or without prepending character names, and with or without prefill.

Doctor-Shotgun/L3.3-70B-Magnum-v5-SFT-Alpha

Text Generation • 71B • Updated May 20 • 7

Note Out of the L3.3 Magnum tunes, this one is the mad and prompt-format-sensitive genius. Works best with prefill + no character names, matching the formatting of the majority of the RP data in the set, can be difficult to wrangle otherwise.

Doctor-Shotgun/L3.3-70B-Magnum-v4-SE

Text Generation • 71B • Updated May 20 • 1.26k • 16

Note An epilogue, if you will, to the v4 series. With minor dataset updates and a new base, this model has been an upgrade to v4 72b in my own testing.

Sao10K/70B-L3.3-Cirrus-x1

Text Generation • 71B • Updated Jan 7 • 1.65k • • 34

Note Sao did it once again - a smart model with a refreshing prose style!