Phi-3.5-MoE Instruct vs Qwen3.5-397B-A17B

Phi-3.5-MoE Instruct

Microsoft

Phi-3.5-MoE-instruct is a sparse mixture-of-experts model from Microsoft's Phi research team, released in August 2024 with 42 billion total parameters across 16 experts and approximately 6.6 billion active parameters per forward pass. The model applies Microsoft's small-data, high-quality training philosophy — developed across earlier Phi generations — to a MoE architecture, targeting reasoning quality comparable to much larger dense models at a fraction of the inference compute. Released under the MIT license, it was notable in the research community for demonstrating that MoE efficiency gains could be realized at smaller total parameter counts than typical large-scale MoE deployments.

Alibaba / Qwen

Qwen3.5-397B-A17B is a 397-billion-parameter mixture-of-experts model from Alibaba's Qwen team, released in February 2026 as the open-weight flagship of the Qwen3.5 series, featuring 17 billion active parameters per forward pass through a hybrid linear-attention and sparse-MoE architecture based on Gated Delta Networks. The model was co-trained on text, images, and video using early fusion, making it natively multimodal across a 262K token context window, while achieving significantly higher inference throughput than comparable dense models due to its sparse computation design. At release it was one of the most capable open-weight models publicly available, offered under Apache 2.0 and accessible through Alibaba's DashScope API as the Qwen3.5-Plus endpoint.