llamaperf

Qwen3.8 125B (6B active) Flash-Next

on NVIDIA RTX 5090 · Strata · 262,144 ctx

NVIDIA hardware
Nov 15, 2026
Throughput
160-195 t/s gen · 6710.2 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
IQ3_S (GGUF)
KV cache
int8
System RAM
120 GB
VRAM reported
32 GB

Summary

User reports Qwen3.8-Flash-Next IQ3_S on an RTX 5090 32GB at 160-195 tok/s decode for real-world single-user workloads, with 195.8 tok/s on an 8059-token cold prompt and 6710.2 tok/s prefill on that same prompt. Setup is Strata v0.1.41 with IQ3_S (84GB) and int8 KV cache at 262144 context, MTP speculation at spec=6, one slot with sequential queuing rather than parallel batching. The ~230 tok/s figure is a hot-cache artifact from a 99.9% KV cache hit rate and is not representative; short-prompt prefill (~200 tok/s) was retracted as an invalid measurement on tiny inputs.

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 NVIDIA RTX 5090

Bars are medians on one scale, 0 to 563.1 tokens/s. Each dot is one run and opens its report. Press ▶ to watch text arrive at a row's speed.

Everything reported on the NVIDIA RTX 5090 →Every Qwen3.8 report →Estimate Qwen3.8 on this card →