Hardware Corner RTX PRO 6000 Blackwell local LLM benchmark page (March 2026)

Per-SKU llama.cpp table, Updated Benchmarks: March 2026. Single RTX PRO 6000 Blackwell (96 GB GDDR7, 1790 GB/s stated on the page). Tables cover Qwen3 8B/14B/32B Q4_K, Gemma4 26B and 31B Q4_K, Qwen3.5 27B Q4_K, Qwen3 30B-A3B Q4_K, Llama 3.3 70B Q4_K, GLM 4.5 Air 106B Q4_K, gpt-oss 120B Q8_0 and MXFP4, Qwen3.5 122B Q4_K, Mistral 123B Q4_K, each at 4k through the largest context that fit. Token-generation column is batch-1 per-request decode; prompt-processing column is prefill at that KV depth. Build hash published only for the gpt-oss 120B Q8_0 and Qwen3.5 122B rows (llama.cpp 8288); other rows publish no build hash.