GGUF models by Theros
103 GGUF models quantized by Theros on Hugging Face. Sizes run from 40M to 14.8B parameters. The longest trained context is 1000K tokens.
About Theros on Hugging Face
Theros has published 103 GGUF models that ggufmodels.com tracks, across families such as Coldbrew, Qwen and Phi. The most downloaded are Qwen2.5 ColdBrew Apex, Q2.5 ColdBrew Forge Orion Test0 and Qwen3 14B VitriousGrey Test0.
The quants Theros publishes most often are Q4_K_M (81), Q5_K_M (17), F16 (4) and Q6_K (2). Their Q4_K_M files range from 4.7 to 9.1 GB. Parameter counts run from 40M to 14.8B, and the longest trained context window is 1000K tokens.
Frequently asked questions
What GGUF models has Theros published?
103 models are listed here, mainly from the Coldbrew, Qwen and Phi families. The most downloaded is Qwen2.5 ColdBrew Apex.
Which quants does Theros publish?
The most common are Q4_K_M (81 models), Q5_K_M (17 models), F16 (4 models) and Q6_K (2 models).