Quanta AI Labs

AMD · MoE Architecture

Instella-MoE

AMD's fully open MoE language model, 16B total and 2.8B active parameters, trained end-to-end on MI300X/MI325X GPUs. Base checkpoint averages 76.7 on standard benchmarks, ahead of Moonlight-16B-A3B and OLMo-3-7B. Research RAIL license; training code is MIT.

Open the source ↗

Similar Language Models releases

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.