NVIDIA DGX Spark (GB10 128GB) (128GB) — Compatible LLMs & Token Throughput
VRAM compatibility and tokens/sec throughput across 28 open-weight models on NVIDIA DGX Spark (GB10 128GB) (Apple Unified Memory, 128GB memory, 273 GB/s bandwidth, 250 TFLOPS).
76 Models × DGX Spark (GB10 128GB)
| Model | Total / Active | Weights (GiB) | 4-bit @ 8K | 4-bit @ 32K | Decode Speed | Compatibility | Full Compatibility Report |
|---|---|---|---|---|---|---|---|
| Qwen3.8 Flash Next (125B-A6B) | 125B / 6B | 70.83 GiB (Q4_K_M) | 71.65 GiB | 72.21 GiB | ~38.3 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen3.8 27B | 27.78B / 27.78B | 16.35 GiB (UD-Q4_K_XL) | 17.56 GiB | 19.06 GiB | ~11.1 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen3.8 2.4T-A95B | 2400B / 95B | 1355.88 GiB (Q4_K_M) | 1357.58 GiB | 1359.74 GiB | ~1 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| Qwen3.6 35B-A3B | 35.95B / 3B | 20.82 GiB (UD-Q4_K_XL) | 21.6 GiB | 22.07 GiB | ~68.6 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen3.5 397B-A17B | 397B / 17B | 224.56 GiB (Q4_K_M) | 225.47 GiB | 226.17 GiB | ~7.9 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| Qwen3.5 122B-A10B | 122B / 10B | 69.19 GiB (Q4_K_M) | 70.03 GiB | 70.59 GiB | ~23.9 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen3.5 35B-A3B | 35.95B / 3B | 20.5 GiB (Q4_K_M) | 21.28 GiB | 21.75 GiB | ~69.5 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen3.5 27B | 27.78B / 27.78B | 16.19 GiB (Q4_K_M) | 17.4 GiB | 18.9 GiB | ~11.3 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen3.5 9B | 9B / 9B | 5.49 GiB (Q4_K_M) | 6.36 GiB | 7.11 GiB | ~32.5 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen3.5 4B | 4B / 4B | 2.51 GiB (Q4_K_M) | 3.37 GiB | 4.12 GiB | ~67.9 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen3 Coder 480B-A35B | 480B / 35B | 271.39 GiB (Q4_K_M) | 273.9 GiB | 279.71 GiB | ~3.5 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| Qwen3 Coder Next 80B-A3B | 79.67B / 3B | 45.2 GiB (Q4_K_M) | 46.02 GiB | 46.58 GiB | ~69.4 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen3 Coder 30B-A3B | 30.53B / 3.3B | 17.28 GiB (Q4_K_M) | 18.59 GiB | 20.84 GiB | ~61 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen3 30B-A3B Instruct 2507 | 30.53B / 3.3B | 17.28 GiB (Q4_K_M) | 18.59 GiB | 20.84 GiB | ~61 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen3 235B-A22B Instruct 2507 | 235.09B / 22B | 132.39 GiB (Q4_K_M) | 134.42 GiB | 138.83 GiB | ~10.2 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| Qwen3 32B | 32.76B / 32.76B | 18.4 GiB (Q4_K_M) | 20.96 GiB | 26.96 GiB | ~9.7 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen3 14B | 14.77B / 14.77B | 8.38 GiB (Q4_K_M) | 10.19 GiB | 13.94 GiB | ~21 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen3 8B | 8.19B / 8.19B | 4.68 GiB (Q4_K_M) | 6.37 GiB | 9.74 GiB | ~36.3 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen3 4B Instruct 2507 | 4.02B / 4.02B | 2.33 GiB (Q4_K_M) | 4.02 GiB | 7.39 GiB | ~67.2 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen3 1.7B | 1.72B / 1.72B | 1.1 GiB (Q4_K_M) | 2.54 GiB | 5.16 GiB | ~126.4 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen3 0.6B | 0.75B / 0.75B | 0.49 GiB (Q4_K_M) | 1.92 GiB | 4.54 GiB | ~216.2 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen2.5 72B Instruct | 72.71B / 72.71B | 41.55 GiB (Q4_K_M) | 44.62 GiB | 52.12 GiB | ~4.4 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Qwen2.5 Coder 32B Instruct | 32.76B / 32.76B | 18.81 GiB (Q4_K_M) | 21.37 GiB | 27.37 GiB | ~9.5 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| QwQ 32B Reasoning | 32.76B / 32.76B | 18.81 GiB (Q4_K_M) | 21.37 GiB | 27.37 GiB | ~9.5 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| DeepSeek V4.1 Flash (552B-A16B) | 552B / 16B | 311.93 GiB (Q4_K_M) | 313.12 GiB | 314.99 GiB | ~7.2 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| DeepSeek V4 Flash | 284B / 11B | 144.44 GiB (UD-Q4_K_XL) | 145.06 GiB | 145.22 GiB | ~19.8 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| DeepSeek V3.2 (685B) | 685.36B / 37B | 377.56 GiB (Q4_K_M) | 378.79 GiB | 380.75 GiB | ~3.1 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| DeepSeek R1-0528 671B | 671B / 37B | 377 GiB (Q4_K_M) | 378.11 GiB | 379.71 GiB | ~3.1 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| DeepSeek V3-0324 671B | 671B / 37B | 377 GiB (Q4_K_M) | 378.11 GiB | 379.71 GiB | ~3.1 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| DeepSeek R1 Distill Llama 70B | 70.55B / 70.55B | 40.25 GiB (Q4_K_M) | 43.32 GiB | 50.82 GiB | ~4.5 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| DeepSeek R1 Distill Qwen 32B | 32.76B / 32.76B | 18.49 GiB (Q4_K_M) | 21.05 GiB | 27.05 GiB | ~9.7 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| DeepSeek R1 Distill Qwen 14B | 14.77B / 14.77B | 8.65 GiB (Q4_K_M) | 10.71 GiB | 15.21 GiB | ~20.2 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| DeepSeek R1 Distill Llama 8B | 8.03B / 8.03B | 4.74 GiB (Q4_K_M) | 6.3 GiB | 9.3 GiB | ~36.1 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| gpt-oss-20b | 20.91B / 3.6B | 10.83 GiB (Q4_K_M) | 11.58 GiB | 12.14 GiB | ~64.9 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| gpt-oss-120b | 116.83B / 5.1B | 58.46 GiB (Q4_K_M) | 59.31 GiB | 60.15 GiB | ~49.2 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Gemma 4 31B IT | 31.27B / 31.27B | 17.07 GiB (Q4_K_M) | 19.04 GiB | 20.91 GiB | ~10.6 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Gemma 4 26B-A4B IT | 25.81B / 3.8B | 15.84 GiB (UD-Q4_K_XL) | 16.75 GiB | 17.22 GiB | ~52.9 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Gemma 4 12B IT | 11.96B / 11.96B | 6.63 GiB (Q4_K_M) | 7.63 GiB | 8 GiB | ~26.9 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Gemma 3 27B IT | 27.43B / 27.43B | 15.41 GiB (Q4_K_M) | 17 GiB | 18.88 GiB | ~11.7 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Gemma 3 12B IT | 12.19B / 12.19B | 7.28 GiB (Q4_K_M) | 8.65 GiB | 10.15 GiB | ~24.3 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Gemma 3 4B IT | 4.3B / 4.3B | 2.7 GiB (Q4_K_M) | 3.56 GiB | 4.12 GiB | ~63.5 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| GLM-5.3 Flash (320B-A18B) | 320B / 18B | 180.93 GiB (Q4_K_M) | 181.84 GiB | 182.9 GiB | ~8.8 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| GLM-5.3 (753B) | 753.33B / 40B | 435.2 GiB (UD-Q4_K_XL) | 436.61 GiB | 439.12 GiB | ~2.7 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| GLM-5 (744B) | 744B / 40B | 420.45 GiB (Q4_K_M) | 421.71 GiB | 423.76 GiB | ~2.8 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| GLM-4.5 (355B-A32B) | 355B / 32B | 200.75 GiB (Q4_K_M) | 204.19 GiB | 212.81 GiB | ~4.5 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| GLM-4.5 Air (110B-A12B) | 110.47B / 12B | 67.96 GiB (Q4_K_M) | 69.96 GiB | 74.27 GiB | ~18 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| GLM-4 32B-0414 | 32B / 32B | 18.44 GiB (Q4_K_M) | 20.92 GiB | 26.64 GiB | ~9.7 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Kimi K3 (2.8T-A104B) | 2780B / 104B | 1570.1 GiB (Q4_K_M) | 1571.49 GiB | 1573.94 GiB | ~0.9 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| Kimi K2 Instruct (1T-A32B) | 1000B / 32B | 565.08 GiB (Q4_K_M) | 566.19 GiB | 567.79 GiB | ~3.2 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| Kimi-Dev 72B | 72.7B / 72.7B | 41.54 GiB (Q4_K_M) | 44.61 GiB | 52.11 GiB | ~4.4 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| MiniMax M3.1 Flash (428B-A23B) | 428B / 23B | 242.14 GiB (Q4_K_M) | 243.65 GiB | 246.46 GiB | ~5.7 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| MiniMax M2.5 (229B-A10B) | 228.69B / 10B | 128.84 GiB (Q4_K_M) | 131.34 GiB | 137.15 GiB | ~21.4 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| Step-3.7 Flash (198B-A11B) | 198B / 11B | 112 GiB (Q4_K_M) | 112.91 GiB | 113.76 GiB | ~21.8 tok/s | Tight Fit (<8% Headroom) | Full Compatibility Report → |
| Step-3.5 Flash (197B-A11B) | 196.81B / 11B | 111.33 GiB (Q4_K_M) | 112.24 GiB | 113.09 GiB | ~21.8 tok/s | Tight Fit (<8% Headroom) | Full Compatibility Report → |
| ERNIE 4.5 300B-A47B (文心) | 300B / 47B | 169.72 GiB (Q4_K_M) | 171.98 GiB | 177.04 GiB | ~3.6 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| ERNIE 4.5 21B-A3B (文心) | 21B / 3B | 11.96 GiB (Q4_K_M) | 12.96 GiB | 14.27 GiB | ~67.8 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Hunyuan A13B Instruct (混元 80B) | 80B / 13B | 45.38 GiB (Q4_K_M) | 46.94 GiB | 49.94 GiB | ~18.2 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| MiniCPM5 2B (面壁) | 2.3B / 2.3B | 1.41 GiB (Q4_K_M) | 2.3 GiB | 3.28 GiB | ~113.1 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Llama 4 Maverick 17B-128E (400B) | 400B / 17.2B | 226.26 GiB (Q4_K_M) | 228.32 GiB | 232.82 GiB | ~7.6 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| Llama 4 Scout 17B-16E (109B) | 109B / 17.2B | 61.95 GiB (Q4_K_M) | 64.01 GiB | 68.51 GiB | ~13.8 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Llama 3.1 405B Instruct | 405B / 405B | 229.51 GiB (Q4_K_M) | 234.05 GiB | 245.86 GiB | ~0.8 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| Llama 3.3 70B Instruct | 70.55B / 70.55B | 39.6 GiB (Q4_K_M) | 42.67 GiB | 50.17 GiB | ~4.6 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Llama 3.1 8B Instruct | 8.03B / 8.03B | 4.58 GiB (Q4_K_M) | 6.14 GiB | 9.14 GiB | ~37.3 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Llama 3.2 3B Instruct | 3.21B / 3.21B | 1.97 GiB (Q4_K_M) | 3.41 GiB | 6.03 GiB | ~79.4 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Mistral Large 4 (1T-A52B) | 1000B / 52B | 565.12 GiB (Q4_K_M) | 566.32 GiB | 568.22 GiB | ~2 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| Mistral Large 2 123B | 122.6B / 122.6B | 69.86 GiB (Q4_K_M) | 73.19 GiB | 81.44 GiB | ~2.6 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Mixtral 8x22B Instruct | 140.6B / 39.1B | 79.46 GiB (Q4_K_M) | 81.78 GiB | 87.03 GiB | ~6.3 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Mixtral 8x7B Instruct | 46.7B / 12.9B | 26.42 GiB (Q4_K_M) | 27.98 GiB | 30.98 GiB | ~18.4 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Mistral Small 3.2 24B | 24.01B / 24.01B | 13.35 GiB (Q4_K_M) | 15.16 GiB | 18.91 GiB | ~13.4 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Devstral Small 24B | 23.6B / 23.6B | 13.59 GiB (Q4_K_M) | 15.4 GiB | 19.15 GiB | ~13.2 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Nemotron 3 Ultra 550B-A55B | 550B / 55B | 310.97 GiB (Q4_K_M) | 311.95 GiB | 312.23 GiB | ~2.2 tok/s | Exceeds Available GPU Memory | Full Compatibility Report → |
| Nemotron 3 Super 120B-A12B | 123.61B / 12B | 78.02 GiB (UD-Q4_K_XL) | 78.81 GiB | 78.99 GiB | ~18.2 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Nemotron 3 Nano 30B-A3B | 31.58B / 3.5B | 22.89 GiB (Q4_K_M) | 23.54 GiB | 23.68 GiB | ~50.3 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Cohere Command A 111B | 111B / 111B | 63.92 GiB (Q4_K_M) | 65.75 GiB | 67.25 GiB | ~2.9 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Phi-4 14B | 14.66B / 14.66B | 8.28 GiB (Q4_K_M) | 10.4 GiB | 15.09 GiB | ~21.1 tok/s | Fits with Safe Headroom | Full Compatibility Report → |
| Liquid LFM3 1.7B | 1.7B / 1.7B | 1.06 GiB (Q4_K_M) | 1.73 GiB | 2.01 GiB | ~150.2 tok/s | Fits with Safe Headroom | Full Compatibility Report → |