More capacity on every GPU
More KV cache memory · 100% model accuracy
Benchmarks
+109%
More KV cache (2.09x). Gemma 4 12B (BF16) on RTX 5090.
+39%
More KV cache (1.39x). Qwen 3.8-27B (BF16) on RTX PRO 6000 Blackwell.
FP8 & FP4
Available for pilots. BF16 is in general release.