Parameters
671B MoE · 37B active · FP8Open Model Catalogue
DeepSeek
Text generationReasoningCoding
DeepSeek V3.2
deepseek-ai/DeepSeek-V3.2DeepSeek flagship chat + reasoning — 8 × H200
Min VRAM
≈ 760 GBHardware
10 × 80 GB GPUOpenAI-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.