community pool / model answer

DeepSeek-V4-Flash-Abliterated

chat · fraserprice/DeepSeek-V4-Flash-Abliterated

199.4 tok/s · model median · 2 runs in the model record.

Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
RTX PRO 6000 Blackwell 96GB ×28-bit181.3 tok/sreported45- GB2,048vllm
RTX PRO 6000 Blackwell 96GB ×48-bit217.5 tok/sreported42- GB2,048vllm

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.