community pool / model answer

Laguna-XS-2.1

chat · poolside/Laguna-XS-2.1

94.1 tok/s · model median · 4 runs in the model record.

Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
Arc Pro B60 24GB4-bit72.5 tok/sreported627- GB2,048llama.cpp
Arc Pro B70 32GB4-bit87.8 tok/sreported491- GB2,048llama.cpp
Radeon AI Pro R9700 32GB ×38-bit100.3 tok/sreported30647.1 GB792llama.cpp
RTX 3090 24GB4-bit296 tok/sreported-- GB2,048lucebox

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.