OpenWeightsTerminal

Qwen3.8-Flash-Next

Qwen3.8-Flash-Next

RTX 3080 10GB · IQ3_S · Strata

Decode · tok/s
52.7
Prefill
—
tok/100W
16.5
Qwen 3.8 Flash IQ3_S (GSQ RCO) on a single RTX 3080 10GB with Strata. 52.7 tok/s last completed response. 9.7 / 10GB VRAM used.

First-person 3080 10GB. Rank is the quoted 52.7. Prefill ~529 stays a note, not a second rank.

Officialclaimed
Open pair