Running Qwen 3.8 27B on GB300 NVL72

Qwen 3.8 27B has 27 billion parameters. GB300 NVL72 offers 20,700 GB of HBM3e across a 72-GPU rack memory. This page puts the two side by side.

By QDNA · Published on 27 July 2026 · Updated on 15 September 2026

How much memory does Qwen 3.8 27B need?

Weights take the parameter count multiplied by the format size, plus a 15% runtime margin for activations and buffers. In FP16 the calculation reads: 27 billion parameters × 2 bytes (FP16) = 54.0 GB of weights; × 1.15 runtime margin = 62.1 GB. The attention cache sits on top: it grows with context length and with the number of concurrent requests, so it is sized case by case.

Model sheet: 27 billion parameters, dense architecture, 262,144-token context, Apache 2.0 licence. Dense model: every parameter computes on each token. The BF16 repository holds 27.8 billion elements (56 GB).

What is left to serve requests?

Qwen 3.8 27B fits comfortably on GB300 NVL72. In FP16, its weights take about 62.1 GB of the 20,700 GB available, leaving 20,638 GB for the attention cache and concurrency. The remaining margin decides how many concurrent requests and how much context the machine sustains. By QDNA convention a machine “fits” when the weights take at most 75% of its memory; beyond that it “barely fits” and forces a limit on context or on concurrency. The attention cache is not calculated on this page. These models’ architectures (compressed latent, linear or sparse attention, Mamba layers) share no per-token formula; it is measured on the machine, with the target context and concurrency [TO BE MEASURED].

GB300 NVL72: price on quotation. NVIDIA publishes no price; analyst estimates range from $3.7M to $6.5M per rack (Loop Capital, Tom’s Hardware, as reported by Spheron on 16 August 2026). The machine targets the GE segment.

Where else can Qwen 3.8 27B run?

• Qwen 3.8 27B on DGX Spark
• Qwen 3.8 27B on Mac Studio Ultra
• Qwen 3.8 27B on DGX Station
• Qwen 3.8 27B on RTX PRO 6000 server
• Qwen 3.8 27B on H200 SXM server

See the Qwen 3.8 27B and GB300 NVL72 fact sheets.

Method and sources

Memory footprints are calculated, not measured: a reading on real hardware may differ depending on the engine and the exact weight format. Every figure below carries its source and the date of its last check.

• Qwen 3.8 27B: 27 billion parameters, 262,144-token context, Apache 2.0 licence. Hugging Face repository (config.json, API, model card), checked on 2 September 2026. Dense model: every parameter computes on each token. The BF16 repository holds 27.8 billion elements (56 GB).
• GB300 NVL72: 20,700 GB of memory; 72 Blackwell Ultra GPUs and 36 Grace CPUs, 20 TB of GPU memory per NVIDIA (72 × 288 GB = 20,736 GB), 37 TB fast memory, 130 TB/s NVLink, 1,440 PFLOPS FP4 with sparsity (1,080 dense); power draw not published. Manufacturer sheet checked on 2 September 2026.
• GB300 NVL72 price: on quotation. NVIDIA publishes no price; analyst estimates range from $3.7M to $6.5M per rack (Loop Capital, Tom’s Hardware, as reported by Spheron on 16 August 2026).
• Bytes per parameter: FP16 2 (16 bits per parameter, hence 2 bytes); FP8 1 (8 bits per parameter (E4M3), hence 1 byte, block scales not counted); NVFP4 0.5 (4 bits per parameter, hence 0.5 byte; 4-bit published weights weigh 0.54 to 0.56 byte per parameter with their scales (DeepSeek V4 Pro 865 GB for 1,599 billion, Kimi K3 1,561 GB for 2,780 billion)); Q4 0.6 (taken here as llama.cpp Q4_K_M, 0.6 byte per parameter including scales; Q4_K_S weighs 0.56, AWQ and GPTQ 0.55). Source: connaissance/faits.yaml, quantifications family, checked on 2026-08-31.
• Runtime margin × 1.15: QDNA operating assumption, not measured: 15% above the weights for activations, buffers and fragmentation.
• “Fits” threshold at 75% of memory: QDNA assumption, not measured, which keeps the remaining quarter for the attention cache and concurrency.