Files
QodeAssist/sources/agents/lmstudio_chat_gemma4_26b_a4b.toml
2026-07-03 09:49:42 +02:00

15 lines
828 B
TOML

schema_version = 1
extends = "LM Studio Chat"
name = "LM Studio Chat — 24 GB (Gemma 4 26B A4B)"
description = "Balanced tier. Local LM Studio coding chat on Gemma 4 26B-A4B QAT (Q4_0) — ~15 GB weights, ~18 GB resident; wants ~24 GB. Mixture-of-experts (~4B active) so it runs at ~12B speed with 26B-class quality — the best all-round pick. Native reasoning + tools + vision. Point the LM Studio (Responses API) provider URL at your host. Note: on LM Studio the context length (KV cache) is a model-load setting, not per-request — the agent's only memory lever is the model."
model = "google/gemma-4-26b-a4b-qat"
enable_thinking = true
tags = ["chat", "lmstudio", "responses", "local", "gemma4", "24gb"]
[body]
temperature = 1.0
top_p = 0.95
max_output_tokens = 16384