Wire · ai
AMD Releases Instella-MoE-16B-A3B: A Fully Open Mixture-of-Experts LLM With 2.8B ...
◆ Sectors
◆ Source
◆ Verified
Fusion42 · 2 August 2026 · Fusion42 review
AMD released Instella-MoE-16B-A3B, an open-source Mixture-of-Experts language model with 16 billion parameters and 2.8 billion active per token, trained on AMD Instinct GPUs. The model is research-focused with open training code under MIT license but commercial use is restricted by the ResearchRAIL license on weights.
This Wire brief sits within Fusion42's coverage of AI & ML.
◆ ◆ The Wire takeaway
You should explore AMD's fully open Mixture-of-Experts model as a new benchmark for research and derivative work, while noting its commercial limits. This opens code-level innovation chances in AI training pipelines but excludes turnkey product deployment.
◆ Coverage
1 source · 2 Aug 2026
◆ Related on Wire
◆ Topics