The LocalLLaMA post links to two models under the GrootTeam account on Hugging Face: Groot-280B-A16B-261023 and Groot-19B-A3B-261023. It describes the 19B-A3B model as a smaller version.
The post claims the smaller model scores higher than Qwen3.6-35B-A3B and Gemma-4-26B-A4B, and produces shorter answers than both. It provides no benchmark details or independent test results.