inclusionAI/Ling-mini-base-2.0

17B • Updated Sep 26 • 153 • 23

Note FAST! Solid function calling, but roleplay takes quite a bit of prompt trial and error. Good context memory.

Orion-zhen/Qwen3-30B-A3B-Instruct-2507-IQK-GGUF

31B • Updated Aug 10 • 227 • 1

Note Excellent roleplay, flawless function calling and logic. Poor memory. Prefill is slow, but cache makes it very fast. Use intelligent expert skipping for faster generation. Requires https://github.com/ikawrakow/ik_llama.cpp

Tiiny/SmallThinker-21BA3B-Instruct

Text Generation • 22B • Updated Jul 31 • 104 • 110

Note Fantastic all-around. Good roleplay, consistent function calling. When prompt caching is enabled, prefill is really fast, but token generation is a bit slow. Initial prefill is painfully slow but very fast after. Q3_K_M, temp 0.6, top_p 0.95, top_k 20.

mradermacher/BitCPM4-1B-GGUF

1B • Updated 19 days ago • 488 • 1

Note Works unusually well for a 1B parameter model, and it's ternary! Doesn't seem to support prompt cache so prefill is slow, but text generation is absurdly fast.