Can DGX Spark (GH200 480GB) run Mistral-Medium-3.5-128B?

NVIDIA · 480 GB LPDDR5X · 500 GB/s bandwidth · Mistral · 128.0B · model ctx up to 256K

CUDA

Yes — the DGX Spark (GH200 480GB) (480 GB LPDDR5X) can run Mistral-Medium-3.5-128B. At Q4 with the default 8K context it needs ≈ 97.0 GB VRAM (weights 70.4 GB + KV cache 25.6 GB + 1 GB runtime overhead), which fits within 480 GB. Decode at Q4/8K: ≈ 2.5 tok/s (estimated, batch 1). FP16 (≈ 282.6 GB) also fits, at ≈ 0.7 tok/s (estimated). Estimates come from memory-bandwidth math; rows tagged measured override estimates. Relative ranking is more reliable than absolute tok/s. Methodology. Buy DGX Spark (GH200 480GB) on Amazon →

Fit & speed by quant and context

Quant	Context	VRAM needed	Fits?	tok/s (decode)
Q4	4K	84.2 GB	Yes	2.5 ± est.
Q4	8K default	97.0 GB	Yes	2.5 ± est.
Q4	32K	173.8 GB	Yes	2.5 ± est.
FP16	4K	269.8 GB	Yes	0.7 ± est.
FP16	8K default	282.6 GB	Yes	0.7 ± est.
FP16	32K	359.4 GB	Yes	0.7 ± est.

Fits = weights + KV(ctx) + 1 GB overhead ≤ 95% of VRAM. Measured anchors are context-agnostic; the fit verdict is recomputed per context. A missing tok/s means the model is far beyond this card (offload-only territory).

VRAM breakdown at 8K context

Quant	Weights	KV cache	Overhead	Total needed	DGX Spark (GH200 480GB) VRAM
Q4	70.4 GB	25.6 GB	1 GB	97.0 GB	480 GB
FP16	256.0 GB	25.6 GB	1 GB	282.6 GB	480 GB

Other GPUs that run Mistral-Medium-3.5-128B

• B200 192GB SXM — ≈39.8 tok/s Q4 est.
• MI300X 192GB — ≈26.3 tok/s Q4 est.
• H200 141GB SXM5 — ≈23.9 tok/s Q4 est.
• Gaudi 3 128GB — ≈16.3 tok/s Q4 est.
• Mac Studio M2 Ultra 192GB — ≈4.0 tok/s Q4 est.
• MacBook Pro M4 Max 128GB — ≈2.7 tok/s Q4 est.
• MacBook Pro M3 Max 128GB — ≈2.0 tok/s Q4 est.

All GPUs for Mistral-Medium-3.5-128B →

Other models for the DGX Spark (GH200 480GB)

• MiniCPM5-1B — ≈294.6 tok/s Q4 est.
• Qwen3.5-2B — ≈159.1 tok/s Q4 est.
• DeepSeek-R1-0528-Qwen3-8B-layer-mix-bpw-3.8-mlx — ≈138.3 tok/s Q4 est.
• Qwen3.6-35B-A3B — ≈106.1 tok/s Q4 est.
• Qwen3.5-35B-A3B — ≈106.1 tok/s Q4 est.
• GLM-4.7-Flash — ≈106.1 tok/s Q4 est.
• Qwen3-30B-A3B-Instruct-2507 — ≈106.1 tok/s Q4 est.
• Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 — ≈106.1 tok/s Q4 est.

All models on the DGX Spark (GH200 480GB) →