Benchmarking Qwen 3.6 35B MoE (3B active) on an RTX 3090

Other: AI & machine learning(gilesthomas.com)view on HackerNews

Qwen3.6-35B-A3BRTX 3090VulkanCUDALlama.cppperformancebenchmarkingAImachine learninglanguage models.

Author: gpjt

Date: 7/24/2026

Article Summary:

A detailed analysis of the performance of the Qwen3.6-35B-A3B model on an RTX 3090 GPU, comparing the Vulkan and CUDA versions of Llama.cpp.