Running gemma4:31b-mlx locally with Ollama feels indistinguishable from paid tiers

walkingriver · x · 2026-09-06

A developer ran gemma4:31b-mlx locally on a MacBook via Ollama and found the answers essentially indistinguishable from his paid subscription, with 100% local inference — the only cost being a spinning fan.

Original post →

More from Infra

Infra channel →