community pool / model answer

Meta-Llama-3-8B-Instruct

chat · meta-llama/Meta-Llama-3-8B-Instruct

48.4 tok/s · model median · 31 runs in the model record.

Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
Arc Pro B60 24GB4-bit43.1 tok/smeasured1,117- GB4,096llama.cpp
Arc Pro B60 24GB ×24-bit46.7 tok/smeasured874- GB4,096llama.cpp
RTX 5080 16GB4-bit129.5 tok/smeasured79- 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.