A large-scale open-weight text model from the Qwen3 family, published by Amazon, that handles long contexts up to 131,072 tokens. It carries the instruct-tuned orientation of the Qwen3 line, meaning it's shaped for following directions and conversational exchanges rather than raw completion tasks. The 'GKA primed' and 'HQwen3' naming suggests a customized or fine-tuned variant, though specific details about what distinguishes this version from base Qwen3 32B are limited.