community pool / model answer

Llama-2-7b-hf

chat · meta-llama/Llama-2-7b-hf

50.4 tok/s · model median · 5 runs in the model record.

Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
Arc Pro B60 24GB2-bit35.3 tok/sreported367- GB4,096llama.cpp
Arc Pro B60 24GB ×22-bit34.8 tok/sreported367- GB4,096llama.cpp
RTX 5080 16GB2-bit201.6 tok/sreported81- GB4,096llama.cpp

Context tested

The table above is the tested context for each exact rig and quantization cell. Empty fields remain empty rather than being filled with an estimate.