Qwen3.8-Flash-Next (5.05bpw + ngram at bf16) exl3 on one r9700: 863 t/s prefill and 35 t/s decode at 230k context (256k max), is that ok or am i missing something?

A user is seeking validation for their Qwen3.8-Flash-Next model setup on a single r9700, using the turboderp exl3 5.05 bpw version. They report 863 t/s prefill and 35 t/s decode at 230k context, with a 256k max. The user is unsure if these performance metrics are optimal or if further tuning is possible for their specific hardware configuration.