← AI Hashrate 中文

Can DGX Spark (GH200 480GB) run MiniMax-M2.7?

NVIDIA · 480 GB LPDDR5X · 500 GB/s bandwidth · MiniMax · 228.7B (active 10B) · MoE · model ctx up to 200K

SS · ExcellentComposite score: VRAM headroom × 0.5 + tok/s speed × 0.5. Q4 @ 8K context.
CUDA

Yes — the DGX Spark (GH200 480GB) (480 GB LPDDR5X) can run MiniMax-M2.7. At Q4 with the default 8K context it needs ≈ 128.79 GB VRAM (weights 125.79 GB + KV cache 2.0 GB + 1 GB runtime overhead), which fits within 480 GB. Decode at Q4/8K: ≈ 31.8 tok/s (estimated, batch 1). FP16 (≈ 460.4 GB) also fits, at ≈ 8.8 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

QuantContextVRAM neededFits?tok/s (decode)
Q44K127.79 GBYes31.8 ± est.
Q48K default128.79 GBYes31.8 ± est.
Q432K134.79 GBYes31.8 ± est.
FP164K459.4 GBYes8.8 ± est.
FP168K default460.4 GBYes8.8 ± est.
FP1632K466.4 GBYes8.8 ± 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

QuantWeightsKV cacheOverheadTotal neededDGX Spark (GH200 480GB) VRAM
Q4125.79 GB2.0 GB1 GB128.79 GB480 GB
FP16457.4 GB2.0 GB1 GB460.4 GB480 GB

Other GPUs that run MiniMax-M2.7

All GPUs for MiniMax-M2.7 →

Other models for the DGX Spark (GH200 480GB)

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