← Library · Frontier
AMD Releases Instella-MoE-16B-A3B-Think, an Open Mixture-of-Experts Model
AMD has introduced Instella-MoE-16B-A3B-Think, an open Mixture-of-Experts (MoE) language model with 16 billion total parameters and 2.8 billion active parameters per token. This model was trained entirely on AMD Instinct MI300X and MI325X GPUs, utilizing AMD's ROCm software stack. It is distributed on Hugging Face in BF16/F32 safetensors format and is currently restricted to academic and research purposes due to its ResearchRAIL license.
Why it matters
This release provides researchers with a powerful, open-source MoE model trained on AMD hardware, fostering innovation and benchmarking within the AI community. Its open nature allows for broader experimentation and development.
Learn one new AI thing every day.
Daily Deck sends you seven plain-English cards like this every morning. Free.
Start free