QriticalBeta
Open Model Catalogue
Alibaba Qwen
Alibaba Qwen
Text generationReasoningVision

Qwen3.5 397B MoE

Qwen/Qwen3.5-397B-A17B-FP8

Qwen frontier — 4 × H200

Deploy Qwen3.5 397B MoE nowView on HuggingFace
Parameters
397B MoE · 17B active · FP8 · vision
Min VRAM
≈ 460 GB
Hardware
6 × 80 GB GPU
OpenAI-compatible API with a per-workload key — drop-in for existing SDKs.
Dedicated GPUs — your model, your instance, no shared queues.
Billed by the second with auto-shutdown — never pay for idle time you didn't plan.