Llama 2 7B on GTX 1070 Ti
43.4best gen tok/s
791best prompt tok/s
640max context tested
4results
Measured · all numbers below come from the linked sources
llama.cpp
| Quant | Gen tok/s | Prompt tok/s | Source | Context tested | TTFT | Build |
|---|---|---|---|---|---|---|
| Q4_0 | 43.4 | 293 | llama.cpp (Vulkan) ↗ | 512+128 | — | llama.cpp Vulkan @860a9e4 — eGPU |
| Q4_0 | 42.9 | 298 | llama.cpp (Vulkan) ↗ | 512+128 | — | llama.cpp Vulkan @860a9e4 — eGPU |
| Q4_0 | 37.9 | 791 | llama.cpp (CUDA) ↗ | 512+128 | — | llama.cpp CUDA @79c1160 |
| Q4_0 | 37.8 | 714 | llama.cpp (CUDA) ↗ | 512+128 | — | llama.cpp CUDA @79c1160 |
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.