OpenWeightsTerminal

Pair

Sign in

OpenWeightsTerminal

Inference Use models Pricing Host & earn

Benchmarks

Models Hardware Compare Market Recipes

Local setup

Get started Get Stacks Runtimes

More data

Pairs Payback Watts Labs Cyber

About

Manifesto How it fits Articles Letter Cite Tape TalkSoon Book Status

Qwen3-8B

M4 Pro · 4bit · rapid-mlx

Decode · tok/s

48.3

Prefill

—

tok/100W

107.3

qwen3-8b → 48.3

Officialclaimed

we <3 open weightsA product of dvidia.orgTalk Read why open weights matter How it fits Terms Privacy Agreementdvidia-inferenceow manifesto litellmv0.5.1 · deffbb8

Home Models Inference Account