GGUF models by mmnga
261 GGUF models quantized by mmnga on Hugging Face. Sizes run from 268M to 671B parameters. The longest trained context is 10240K tokens.
About mmnga on Hugging Face
mmnga has published 261 GGUF models that ggufmodels.com tracks, across families such as Llama, Qwen and Llm. The most downloaded are tokyotech llm Swallow MS 7b instruct v0.1, Llama-3-ELYZA-JP-8B and Llama 3.1 Swallow 8B Instruct v0.5.
The quants mmnga publishes most often are Q4_0 (243), Q3_K_M (240), Q4_K_M (239) and Q3_K_L (238). Their Q4_K_M files range from 0.3 to 404.4 GB. Parameter counts run from 268M to 671B, and the longest trained context window is 10240K tokens.
Frequently asked questions
What GGUF models has mmnga published?
261 models are listed here, mainly from the Llama, Qwen and Llm families. The most downloaded is tokyotech llm Swallow MS 7b instruct v0.1.
Which quants does mmnga publish?
The most common are Q4_0 (243 models), Q3_K_M (240 models), Q4_K_M (239 models) and Q3_K_L (238 models).