RTX 5070 Ti × Meta Llama 3.1 8B Instruct
Meta Llama 3.1 8B Instruct on RTX 5070 Ti
91.8best gen tok/s
4,323best prompt tok/s
4,352max context tested
10results
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 | 91.8 | 3,860 | LocalScore ↗ | 512+128 | 314 ms | llamafile 0.9.2 |
| Q4_K - Medium | 79.8 | 4,323 | LocalScore ↗ | 512+128 | 315 ms | llamafile 0.9.2 |
| Q4_K - Medium | 48.6 | 3,827 | LocalScore ↗ | 512+128 | 335 ms | llamafile 0.9.2 |
| Q4_K - Medium | 48.6 | 4,023 | LocalScore ↗ | 512+128 | 313 ms | llamafile 0.9.3 |
| Q4_K - Medium | 41.9 | 2,617 | LocalScore ↗ | 512+128 | 488 ms | llamafile 0.9.2 |
| Q4_K - Medium | 35.4 | 1,961 | LocalScore ↗ | 512+128 | 646 ms | llamafile 0.9.3 |
| Q4_K - Medium | 33.7 | 1,951 | LocalScore ↗ | 512+128 | 649 ms | llamafile 0.9.3 |
| Q4_K - Medium | 33.3 | 1,955 | LocalScore ↗ | 512+128 | 649 ms | llamafile 0.9.3 |
| Q4_K - Medium | 33.1 | 2,475 | LocalScore ↗ | 512+128 | 503 ms | llamafile 0.9.3 |
| Q4_K - Medium | 24.0 | 3,083 | LocalScore ↗ | 4,352 | 410 ms | llamafile 0.9.3 |
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 RTX 5070 Ti 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.