community pool / model answer

InternVL3-14B

chat · OpenGVLab/InternVL3-14B

32.7 tok/s · model median · 24 runs in the model record.

Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
RTX 3060 12GB ×24-bit30.3 tok/smeasured-0.1 GB65,536llama.cpp
RTX 3090 24GB4-bit43.5 tok/smeasured2,970- 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.