community pool / model answer

Qwen2.5-Coder-7B-Instruct

code · Qwen/Qwen2.5-Coder-7B-Instruct

112.5 tok/s · model median · 6 runs in the model record.

Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
RTX 2080 8GB4-bit93 tok/sreported-4.7 GB2,048ollama
RTX 3060 12GB4-bit119 tok/sreported-4.7 GB2,048ollama
RTX 3080 12GB4-bit140.8 tok/smeasured-8.2 GB2,048llama.cpp, ollama
RTX3060+RTX2080-pooled-20GB 20GB4-bit106 tok/sreported-4.7 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.