32 measured results from 8 builds, each linked to its source.
| Model | Build | Ran on | Quant | Engine | tok/s | Details |
|---|
| DeepSeek R1 Distill 14B | Dell Pro Max with GB10 (Nvidia DGX Spark) | DGX Spark (GB10) | | llama.cpp | 23.92 | |
| GLM-5.2 | 4x DGX Spark cluster running GLM-5.2 (unpruned) | 4x DGX Spark (GB10) | QuantTrio Int4-Int8Mix | vLLM TP4 + MTP, single-stream mean | 32.5 | ⚡ spec-decode (MTP)context: 200K max📏 Single-stream mean (source config of 2026-07-17) |
| GLM-5.3 Flash | 4x DGX Spark ring running GLM-5.3 Flash NVFP4 | 4x DGX Spark (GB10) | NVFP4 | vLLM TP4 (RiNGSiDE), single-stream code | 105.3 | 📏 Single-stream decode reported by the source |
| GLM-5.3 Flash | 4x DGX Spark ring running GLM-5.3 Flash NVFP4 | 4x DGX Spark (GB10) | NVFP4 | vLLM TP4 (RiNGSiDE), single-stream prose | 61.7 | 📏 Single-stream decode reported by the source |
| GLM-5.3 Flash | 2x DGX Spark running GLM-5.3 Flash NVFP4 | 2x DGX Spark (GB10) | NVFP4 | vLLM TP2 (RiNGSiDE), single-stream code | 56.5 | 📏 Single-stream decode reported by the source |
| GLM-5.3 Flash | 2x DGX Spark running GLM-5.3 Flash NVFP4 | 2x DGX Spark (GB10) | NVFP4 | vLLM TP2 (RiNGSiDE), single-stream prose | 33.0 | 📏 Single-stream decode reported by the source |
| Gemma 4 26B-A4B | DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s | DGX Spark (GB10) | + official MTP drafter (k=3) | vLLM 0.31.0, 128K ctx, 1 user | 56.5 | ⚡ spec-decode (MTP) |
| Gemma 4 26B-A4B | DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s | DGX Spark (GB10) | no speculative decoding | vLLM 0.31.0, 128K ctx, 1 user | 30.8 | |
| Llama 3.1 70B | Dell Pro Max with GB10 (Nvidia DGX Spark) | DGX Spark (GB10) | | llama.cpp | 4.71 | |
| Llama 3.1 8B | DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s | DGX Spark (GB10) | | NVIDIA NIM 2.0.9, 128K ctx, 1 user | 26.4 | |
| Llama 3.2 3B | Dell Pro Max with GB10 (Nvidia DGX Spark) | DGX Spark (GB10) | | llama.cpp | 97.93 | |
| Nemotron 3 Nano Omni 30B | DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s | DGX Spark (GB10) | | vLLM 0.31.0, 128K ctx, 1 user | 58.8 | |
| Nemotron 3 Super 120B-A12B | DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s | DGX Spark (GB10) | + MTP (k=3) | vLLM 0.31.0, 128K ctx, 1 user | 23.0 | ⚡ spec-decode (MTP) |
| Nemotron 3 Super 120B-A12B | DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s | DGX Spark (GB10) | no speculative decoding | vLLM 0.31.0, 128K ctx, 1 user | 16.1 | |
| Nemotron 3.5 Bolt 74B-A9B | DGX Spark (GB10, 128GB): Nemotron 3.5 Super VL 120B and Bolt 74B, launch-week runs | DGX Spark (GB10) | 2-bit 24GB GGUF + DFlash speculative decoding (7 draft tokens) | llama.cpp | 61.61 | ⚡ spec-decode (DFlash) |
| Nemotron 3.5 Bolt 74B-A9B | DGX Spark (GB10, 128GB): Nemotron 3.5 Super VL 120B and Bolt 74B, launch-week runs | DGX Spark (GB10) | 2-bit 24GB GGUF, no speculative decoding | llama.cpp | 39.47 | |
| Nemotron 3.5 Lightning 30B-A3B | DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s | DGX Spark (GB10) | + DSpark speculative decoding (k=3) | vLLM 0.31.0, 128K ctx, 1 user | 94.4 | ⚡ spec-decode (DSpark) |
| Nemotron 3.5 Lightning 30B-A3B | DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s | DGX Spark (GB10) | no speculative decoding | vLLM 0.31.0, 128K ctx, 1 user | 70.3 | |
| Nemotron 3.5 Super VL 120B-A12B | DGX Spark (GB10, 128GB): Nemotron 3.5 Super VL 120B and Bolt 74B, launch-week runs | DGX Spark (GB10) | NVFP4 + LiLiCorr speculative decoding (7 draft tokens) | vLLM | 38.07 | ⚡ spec-decode (LiLiCorr) |
| Nemotron 3.5 Super VL 120B-A12B | DGX Spark (GB10, 128GB): Nemotron 3.5 Super VL 120B and Bolt 74B, launch-week runs | DGX Spark (GB10) | NVFP4, no speculative decoding, 256-token output | vLLM | 19.06 | |
| Qwen3.6 27B | DGX Spark with Qwen3.6 27B NVFP4 + DFlash speculative decoding | DGX Spark (GB10) | NVFP4 | vLLM + DFlash spec decoding | 32.83 | ⚡ spec-decode (DFlash)📏 Median decode, Docker reproduction |
| Qwen3.6 27B | DGX Spark with Qwen3.6 27B NVFP4 + DFlash speculative decoding | DGX Spark (GB10) | FP8 | vLLM autoregressive baseline | 7.85 | 📏 Autoregressive baseline, same reproduction |
| Qwen3.6 35B-A3B | DGX Spark (GB10): Qwen3.6 35B-A3B NVFP4 at 113 tok/s on vLLM with MTP | DGX Spark (GB10) | NVFP4 | vLLM 0.30 + MTP | 113 | ⚡ spec-decode (MTP) |
| Qwen3.6 35B-A3B | DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s | DGX Spark (GB10) | NVFP4 + MTP (k=2) | vLLM 0.31.0, 128K ctx, 1 user | 102.3 | ⚡ spec-decode (MTP) |
| Qwen3.6 35B-A3B | DGX Spark (GB10): Qwen3.6 35B-A3B NVFP4 at 113 tok/s on vLLM with MTP | DGX Spark (GB10) | NVFP4 | vLLM 0.30 + MTP | 86 | ⚡ spec-decode (MTP) |
| Qwen3.6 35B-A3B | DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s | DGX Spark (GB10) | NVFP4, no speculative decoding | vLLM 0.31.0, 128K ctx, 1 user | 78.1 | |
| Qwen3.8 27B | DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s | DGX Spark (GB10) | + MTP (k=2) | vLLM 0.31.0, 128K ctx, 1 user | 19.0 | ⚡ spec-decode (MTP) |
| Qwen3.8 27B | DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s | DGX Spark (GB10) | no speculative decoding | vLLM 0.31.0, 128K ctx, 1 user | 11.0 | |
| Qwen3.8 Flash-Next | DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s | DGX Spark (GB10) | + MTP (k=2) | patched vLLM (qwen38-flash-dgx), 128K ctx, 1 user | 26.8 | ⚡ spec-decode (MTP) |
| gpt-oss 120B | DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s | DGX Spark (GB10) | | vLLM 0.31.0, 128K ctx, 1 user | 44.4 | |
| gpt-oss 20B | DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s | DGX Spark (GB10) | | vLLM 0.31.0, 128K ctx, 1 user | 96.2 | |
| gpt-oss 20B | Dell Pro Max with GB10 (Nvidia DGX Spark) | DGX Spark (GB10) | | llama.cpp | 87.97 | |
The fastest plain-decode result we have is 105.3 tok/s on GLM-5.3 Flash (NVFP4, vLLM TP4 (RiNGSiDE), single-stream code) in 4x DGX Spark ring running GLM-5.3 Flash NVFP4. Speed depends heavily on the model, quant and engine, so compare like with like in the table. RigBase lists 32 measured results from 8 builds.
128 GB unified memory per GB10 system.
DeepSeek R1 Distill 14B, GLM-5.2, GLM-5.3 Flash, Gemma 4 26B-A4B, Llama 3.1 70B, Llama 3.1 8B, Llama 3.2 3B, Nemotron 3 Nano Omni 30B, Nemotron 3 Super 120B-A12B, Nemotron 3.5 Bolt 74B-A9B, Nemotron 3.5 Lightning 30B-A3B, Nemotron 3.5 Super VL 120B-A12B, Qwen3.6 27B, Qwen3.6 35B-A3B, Qwen3.8 27B, Qwen3.8 Flash-Next, gpt-oss 120B, gpt-oss 20B.
Every row links to a build page with its original public source (benchmark logs, READMEs, forum posts). RigBase does not re-run the benchmarks. Results marked ⚡ use speculative decoding and are not comparable 1:1 with plain decode.