community pool / model answer

Olmo-3-32B-Think

chat · allenai/Olmo-3-32B-Think

16.4 tok/s · model median · 17 runs in the model record.

Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
RTX 3060 12GB ×24-bit15.4 tok/smeasured10,91321 GB65,536llama.cpp
RTX 3090 24GB4-bit22.2 tok/smeasured9,216- GB131,072llama.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.