community pool / model answer

Llama-3.1-8B-Instruct

chat · meta-llama/Llama-3.1-8B-Instruct

76 tok/s · model median · 8 runs in the model record.

Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
GTX 1080 Ti 11GB4-bit52.8 tok/sreported-7.7 GB8,192llama.cpp
GTX 1080 Ti 11GB5-bit38.1 tok/sreported-8.4 GB8,192llama.cpp
RTX 2080 8GB4-bit76 tok/sreported-4.9 GB2,048ollama
RTX 3060 12GB4-bit76 tok/sreported-4.9 GB2,048ollama
RTX 3080 12GB4-bit134 tok/sreported-5.1 GB2,048ollama
RTX3060+RTX2080-pooled-20GB 20GB4-bit105 tok/sreported-5 GB2,048ollama

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.