schema_version = 1

extends     = "LM Studio Compression"
name        = "LM Studio Compression — 24 GB (Gemma 4 26B A4B)"
description = "Balanced tier. Conversation summarization on Gemma 4 26B-A4B QAT (Q4_0) — ~18 GB resident, 24 GB. Mixture-of-experts (~4B active) so it summarizes at ~12B speed with more capacity. Reasoning disabled via reasoning.effort=none for a direct summary. Point the LM Studio (Responses API) provider URL at your host."

model = "google/gemma-4-26b-a4b-qat"
tags  = ["compression", "lmstudio", "responses", "local", "gemma4", "24gb"]

[body]
reasoning = { effort = "none" }