llamaperf
Nov 15, 2026
Throughput
25.0 t/s gen · 700.0 tokens/s prompt processing (prefill)Prompt processing is how fast the input is read, not how long you wait for the first word. How to measure time to first token.
Quant
TQ2_0 (GGUF)
System RAM
32 GB
VRAM reported
16 GB

Summary

User reports Qwen3.8-Flash-Next at 25 t/s decode and 700 t/s prefill on an RX 6800 XT 16GB. Setup is Strata with a TQ2_0 quant on a Ryzen 7 5800X3D with 32GB DDR4. User asks for a second opinion on the configuration, noting most benchmarks they have seen are Nvidia with very different settings.

Compare this run

Medians of plain runs: one device, one request, no speculative decoding, the whole model in the device's memory. The report above is one setup, so read it beside them and not as one of them.

Qwen3.8 125B · 6B active at 4-bit by hardware

Bars are medians on one scale, 0 to 232.0 tokens/s. Each dot is one run and opens its report. Press ▶ to watch text arrive at a row's speed.

Models most run on the AMD RX 6800 XT 16GB

Bars are medians on one scale, 0 to 11.4 tokens/s. Press ▶ to watch text arrive at a row's speed.

Everything reported on the AMD RX 6800 XT 16GB →Every Qwen3.8 report →Estimate Qwen3.8 on this card →