community pool / model answer

Qwen2.5-7B-Instruct

chat · Qwen/Qwen2.5-7B-Instruct

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

Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
GTX 1080 Ti 11GB4-bit57.6 tok/sreported-7.4 GB8,192llama.cpp
GTX 1080 Ti 11GB5-bit41.1 tok/sreported-8.1 GB8,192llama.cpp
M1 Max 64GB4-bit48.7 tok/sreported30- GB2,048ollama
RTX 4070 SUPER 12GB4-bit77.4 tok/sreported2487 GB65,536llama.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.