Llama 2 7B on RTX 3080
146best gen tok/s
5,570best 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 | 146 | 4,914 | llama.cpp (Vulkan) ↗ | 512+128 | — | llama.cpp Vulkan @7c7d6ce — coopmat2 |
| Q4_0 | 140 | 5,570 | llama.cpp (CUDA) ↗ | 512+128 | — | llama.cpp CUDA @9c35706 |
| Q4_0 | 140 | 5,014 | llama.cpp (CUDA) ↗ | 512+128 | — | llama.cpp CUDA @9c35706 |
| Q4_0 | 139 | 4,287 | llama.cpp (Vulkan) ↗ | 512+128 | — | llama.cpp Vulkan @7c7d6ce — coopmat2 |
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.