community pool / model answer

DeepSeek-Coder-V2-Lite-Instruct

code · deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct

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

Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
RTX 2080 8GB4-bit39 tok/sreported-8.9 GB2,048ollama
RTX 3060 12GB4-bit149 tok/sreported-8.9 GB2,048ollama
RTX 3080 12GB4-bit87 tok/sreported-9.6 GB2,048ollama
RTX3060+RTX2080-pooled-20GB 20GB4-bit118.5 tok/sreported-9.3 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.