AMD Releases Instella-MoE-16B-A3B: A Fully Open Mixture-of-Experts LLM With 2.8B Active Parameters Trained On Instinct GPUs
AMD released Instella-MoE-16B-A3B, a fully open Mixture-of-Experts language model trained from scratch on Instinct MI300X and MI325X GPUs. The model holds 16B total parameters but activates only 2.8B per token. AMD is publishing weights from every training stage, along with data mixtures, training configs, and inference code. Two systems-level choices carry the release: Gated Multi-head Latent Attention and FarSkip-Collective connectivity. Is it deployable? Partly. The weights ship under
Make your prediction
Will AMD release full training configs for Instella-MoE-16B-A3B on GitHub by August 16, 2026?
Resolves by Aug 16, 2026
Your prediction
50% · 50/50 coin flip
NOYES
Get smart on it
AMD released a fully open Mixture-of-Experts language model with 16 billion total parameters but only 2.8 billion active per token, trained on AMD's Instinct GPUs. The model's weights are available under a research-only license, while the training code is MIT-licensed for broader reuse. The release includes two technical innovations: Gated Multi-head Latent Attention and FarSkip-Collective connectivity, which together provide faster training and reduced serving time. The model is designed for AI research labs, universities, and enterprise teams with data-center GPU capacity rather than commercial deployment, and it performs competitively with other open models on standard benchmarks.
Related news
Pokee AI Releases Pokee-Isaac 28B: A 10M-Token Context Agentic Model Built to Run Inside the Customer Boundary
Long-horizon agents accumulate context faster than they resolve tasks. Every tool output, observation, and intermediate reasoning step stays in the window, and the two capabilities that matter — holding that context and staying coherent across it — have so far been available almost exclusively from cloud endpoints. That excludes regulated industries, public-sector institutions, and on-device applications, where the data is not permitted to leave the boundary at all. Pokee AI released Pokee-Isaa
Mistral AI Releases Shieldstral 1.0 3B: An Open-Weights Policy-Adaptive Multimodal Safety Classifier Matching Models 7× Its Size
Mistral AI has released Shieldstral 1.0 3B, an open-weights, policy-adaptive multimodal safety classifier that treats content moderation as a single yes/no question rather than a fixed taxonomy of harm categories. Most guardrail models bake their category list into the weights, so re-targeting one to a new deployment context means retraining — and the same content can be acceptable on a cybersecurity research tool while being harmful on a mental-health platform. Shieldstral inverts that: operat
