RTX 3090 Ti
27 records across 6 models (quants: MXFP4, Q4_K, Q8_0).
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| RTX 3090 Ti | gpt-oss 20B 20B | MXFP4 | llama.cpp | 160.3 | 4805.4 | – | – | 4096 | – | sourced | – | source |
| RTX 3090 Ti | gpt-oss 20B 20B | MXFP4 | llama.cpp | 137.6 | 3440.1 | – | – | 16384 | – | sourced | – | source |
| RTX 3090 Ti | gpt-oss 20B 20B | MXFP4 | llama.cpp | 119.7 | 2521.3 | – | – | 32768 | – | sourced | – | source |
| RTX 3090 Ti | gpt-oss 20B 20B | MXFP4 | llama.cpp | 94.9 | 1666.5 | – | – | 65536 | – | sourced | – | source |
| RTX 3090 Ti | gpt-oss 20B 20B | MXFP4 | llama.cpp | 61.5 | 936.4 | – | – | 131072 | – | sourced | – | source |
| RTX 3090 Ti | Qwen 3.5 9B | Q8_0 | llama.cpp (CUDA) | 81.58 | – | – | – | 32768 | 2026-04-01 | community | – | source |
| RTX 3090 Ti | Qwen 3.5 9B | Q8_0 | llama.cpp (CUDA) | 77.44 | – | – | – | 32768 | 2026-04-01 | community | – | source |
| RTX 3090 Ti | Qwen 3.5 9B | Q8_0 | llama.cpp (CUDA) | 75.55 | – | – | – | 32768 | 2026-04-01 | community | – | source |
| RTX 3090 Ti | Qwen 3.5 9B | Q8_0 | llama.cpp (CUDA) | 75.01 | – | – | – | 32768 | 2026-04-01 | community | – | source |
| RTX 3090 Ti | Qwen 3.5 9B | Q8_0 | llama.cpp (CUDA) | 74.85 | – | – | – | 32768 | 2026-04-01 | community | – | source |
| RTX 3090 Ti | Qwen 3.5 9B | Q8_0 | llama.cpp (CUDA) | 72.79 | – | – | – | 65536 | 2026-04-01 | community | – | source |
| RTX 3090 Ti | Qwen 3.5 9B | Q8_0 | llama.cpp (CUDA) | 63.44 | – | – | – | 65536 | 2026-04-01 | community | – | source |
| RTX 3090 Ti | Qwen 3.5 9B | Q8_0 | llama.cpp (CUDA) | 61.47 | – | – | – | 65536 | 2026-04-01 | community | – | source |
| RTX 3090 Ti | Qwen3 14B 14B | Q4_K | llama.cpp | 76.2 | 2817.1 | – | – | 4096 | – | sourced | – | source |
| RTX 3090 Ti | Qwen3 14B 14B | Q4_K | llama.cpp | 56.9 | 1914.8 | – | – | 16384 | – | sourced | – | source |
| RTX 3090 Ti | Qwen3 14B 14B | Q4_K | llama.cpp | 42.1 | 1232.1 | – | – | 32768 | – | sourced | – | source |
| RTX 3090 Ti | Qwen3 14B 14B | Q4_K | llama.cpp | 28.2 | 828.5 | – | – | 65536 | – | sourced | – | source |
| RTX 3090 Ti | Qwen3 30B A3B 30B | Q4_K | llama.cpp | 166.9 | 3441.0 | – | – | 4096 | – | sourced | – | source |
| RTX 3090 Ti | Qwen3 30B A3B 30B | Q4_K | llama.cpp | 121.9 | 2205.6 | – | – | 16384 | – | sourced | – | source |
| RTX 3090 Ti | Qwen3 30B A3B 30B | Q4_K | llama.cpp | 92.0 | 1483.9 | – | – | 32768 | – | sourced | – | source |
| RTX 3090 Ti | Qwen3 32B 32B | Q4_K | llama.cpp | 38.0 | 1238.8 | – | – | 4096 | – | sourced | – | source |
| RTX 3090 Ti | Qwen3 32B 32B | Q4_K | llama.cpp | 32.6 | 862.7 | – | – | 16384 | – | sourced | – | source |
| RTX 3090 Ti | Qwen3 8B 8B | Q4_K | llama.cpp | 123.7 | 4467.9 | – | – | 4096 | – | sourced | – | source |
| RTX 3090 Ti | Qwen3 8B 8B | Q4_K | llama.cpp | 93.6 | 2834.2 | – | – | 16384 | – | sourced | – | source |
| RTX 3090 Ti | Qwen3 8B 8B | Q4_K | llama.cpp | 71.8 | 1868.0 | – | – | 32768 | – | sourced | – | source |
| RTX 3090 Ti | Qwen3 8B 8B | Q4_K | llama.cpp | 49.0 | 1111.1 | – | – | 65536 | – | sourced | – | source |
| RTX 3090 Ti | Qwen3 8B 8B | Q4_K | llama.cpp | 29.0 | 612.4 | – | – | 131072 | – | sourced | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085