DGX Spark × Llama 3.2 1B Instruct
Llama 3.2 1B Instruct on DGX Spark
158best gen tok/s
8,014best prompt tok/s
4,352max context tested
11results
Measured · all numbers below come from the linked sources
llama.cpp
| Quant | Gen tok/s | Prompt tok/s | Source | Context tested | TTFT | Build |
|---|---|---|---|---|---|---|
| Q4_K - Medium | 158 | 7,964 | LocalScore ↗ | 4,352 | 193 ms | llamafile 0.9.2 |
| Q4_K - Medium | 158 | 7,437 | LocalScore ↗ | 512+128 | 201 ms | llamafile 0.9.2 |
| Q4_K - Medium | 158 | 8,014 | LocalScore ↗ | 4,352 | 193 ms | llamafile 0.9.2 |
| Q4_K - Medium | 153 | 7,578 | LocalScore ↗ | 512+128 | 207 ms | llamafile 0.9.2 |
| Q4_K - Medium | 139 | 7,083 | LocalScore ↗ | 512+128 | 218 ms | llamafile 0.9.2 |
| Q4_K - Medium | 135 | 6,980 | LocalScore ↗ | 512+128 | 222 ms | llamafile 0.9.2 |
| Q4_K - Medium | 118 | 7,480 | LocalScore ↗ | 512+128 | 204 ms | llamafile 0.9.2 |
| Q4_K - Medium | 112 | 7,529 | LocalScore ↗ | 512+128 | 205 ms | llamafile 0.9.2 |
| Q4_K - Medium | 105 | 7,346 | LocalScore ↗ | 512+128 | 208 ms | llamafile 0.9.2 |
| Q4_K - Medium | 103 | 5,148 | LocalScore ↗ | 512+128 | 305 ms | llamafile 0.9.2 |
| Q4_K - Medium | 100 | 5,436 | LocalScore ↗ | 512+128 | 284 ms | llamafile 0.9.2 |
Ollama
No measured Ollama results yet. Submit yours →
MLX
No measured MLX results yet. Submit yours →
"Context tested" is the largest prompt+generation length actually benchmarked (tokens), not a memory-fit estimate.
What else can DGX Spark run? →
Includes results from LocalScore, a Mozilla Builders project by cjpais, mirrored in the cn0303/local-llm-speed-benchmarks dataset. LocalScore runs use llamafile, which is built on llama.cpp.