Qwen3.8 Flash-Next IQ3_S
Token generation is capped by how fast weights stream from memory. Qwen3.8 Flash-Next is a mixture-of-experts model: only 6B of its 180B parameters are read per token, about 2.8 GB at IQ3_S. Layers are split across mixed cards; the ceiling uses the fastest card as an upper bound: 1493 GB/s ÷ 2.8 GB ≈ 534 tok/s max.
Detected by the CLI. The owner can't edit these fields.
| Model | Quant and context | Generation | Prompt processing | Share of the ceiling | Rank | Status |
|---|---|---|---|---|---|---|
| Qwen3.8 Flash-Nexton Strata | IQ3_S8k context | 99.0tok/s | 1,321 tok/s | 19% | #1 | Verified run |
Load Qwen3.8 Flash-Next at IQ3_S and run the command. If your rig has the same GPUs and lands within 10%, this result gains a reproduction.
npx toks-gg benchClosest memory bandwidth, on Qwen3.8 Flash-Next IQ3_S.