DMindAI/DMind-2-107B
#1
by
blackcat1402
- opened
[MODEL_HF_LINK_HERE] This is GLM4.5 Air based fine tuned LLM for crypto purpose. Based on exl3 characteristics, 2.99bpw-optim, 3.0bpw, 3.14bpw may exhibit the best performance for agentic purpose.
Queued.
Thanks for your great work. This fine tuned model seems to be only solid at higher bit quantized version for lower bit quant, its performance dropped severely.