Hardware Corner Mar 2026: RTX 6000 Ada Qwen3 14B (Q4_K) prefill @ 32k ctx
1310.2 tok/s (prefill, per-request, KV depth 32k)
Fields
id
benchmarks/hc26-6000ada-qwen3-14b-pp-32kMetric
prefill_tok_s
Value
1,310.2
Unit
tok/s (prefill, per-request, KV depth 32k)
Model
Qwen3 14B (Q4_K)
Methodology
THIRD-PARTY LAB MEASURED (Hardware Corner per-GPU LLM benchmark page, Updated Benchmarks: March 2026). llama.cpp, single RTX 6000 Ada (48 GB GDDR6 ECC). Context length is the KV depth the table column names. Token Generation column = batch-1 per-request decode tok/s, NOT aggregate across concurrent requests. Prompt Processing column = prefill tok/s at that context depth. The page does not publish the llama.cpp build hash, CUDA version, or exact prompt and output token counts beyond the context-depth bucket.
Measured by
third_party
Reproducible
yes
Thermal state
unknown
notes
format_id is the nearest standard GGUF in this repo; the page labels the quant as shown in the model field. Single card, weights fully in VRAM. A dash in the source table means the context did not fit and is not recorded.
Notes
format_id is the nearest standard GGUF in this repo; the page labels the quant as shown in the model field. Single card, weights fully in VRAM. A dash in the source table means the context did not fit and is not recorded.