Gemma 4 12B Instruct

Google llm

Mid-size Gemma 4 model with an encoder-free unified decoder that ingests raw image and audio patches directly, aimed at local agentic use.

overall ⌄

43.2

52 of 93 ranked

price · $/M tokens

self-hosted

open weights, no first-party API

to run it yourself ⌄

7 GB

minimum · 29 GB recommended

Benchmarks

What its maker published, and what anyone else measured. Every figure links the document it came from.

benchmark	published	measured
MMLU-Pro	77.2	—
GPQA	78.8	—

No recorded lineage — root or standalone model.