Can I run this LLM?
Can a NVIDIA GH200 Grace Hopper (144GB HBM + 480GB) run DeepSeek V4 Flash (MoE)?
DeepSeek · MIT · estimated for a 144.0 GB machine at 4K context.
Yes — it runs great.
Runs greatBest fit: DeepSeek V4 Flash (MoE) at Q4_K_M · 155.1 GB weights · 120–179 tok/s on a NVIDIA GH200 Grace Hopper (144GB HBM + 480GB).
Arena Text Overall · style-controlled
UnratedAmbiguous matchAmbiguous Arena identity; unrated
Arena identity is ambiguous. The local 0731 release is not bound to the unversioned Arena entry or preview configurations. No rating is assigned or used for ranking. Local quantization quality is unmeasured.
Snapshot published 2026-10-08 · snapshot imported 2026-10-11 UTC · validated last-good data · manual refresh
Arena leaderboard dataset · Arena / lmarena-ai, Arena Leaderboard Dataset (CC BY 4.0). Selected Text Overall style-controlled rows; field names adapted, ratings unchanged. · dataset CC BY 4.0 · revision e16f90232986
- Total memory at 4K context
- 156.3 GB
- Weights + KV cache + runtime overhead
- Estimated generation
- 120–179 tok/s
- Estimated first token
- ~0.6s
Every quantization on this machine
On small screens, each quant lists every metric below its name. On wider screens, scroll sideways for total memory, speed and fit. Keyboard: focus the table, then use the arrow keys. Quant names stay visible.
| Quant | Weights | Total memory | Speed | Fit |
|---|---|---|---|---|
| Q2_K | 96.8 GB | 98.0 GB | 252–378 tok/s | Runs great |
| Q3_K_M | 128.1 GB | 129.2 GB | 230–346 tok/s | Runs great |
| Q4_K_M | 155.1 GB | 156.3 GB | 120–179 tok/s | Runs great |
Ran DeepSeek V4 Flash (MoE) on a NVIDIA GH200 Grace Hopper (144GB HBM + 480GB)? Add your real numbers →