Pair

Qwen3.8-Flash-Next

Name: OpenWeights Terminal measured pairs
Creator: Dvidia
License: https://owterminal.com/cite

One quoted capture. Rank only lives inside a filter.

RTX 5070 Ti 16GB · IQ3_S · Strata

Decode · tok/s

53.5

Prefill

—

tok/100W

17.8

That moved cold decode to 43 tok/s, while cached turns reached 53.5 tok/s.

First-person 5070 Ti 16GB. Rank is the quoted 53.5 cached. Single runs, author flagged salt.

Officialclaimed