Skip to content
Hacker News front page

Gemma 4 12B: A Unified, Encoder-Free Multimodal Model

Gemma 4 12B: A unified, encoder-free multimodal model

Google’s title introduces Gemma 4 12B as a unified, encoder-free multimodal model; the RSS snippet only lists 137 Hacker News points and 48 comments, and the post does not disclose architecture details, training setup, pricing, release terms, or benchmark results.

Why it matters: Google released Gemma 4 12B with a unified multimodal design that drops the separate vision encoder, letting one model handle text and images directly.

Read the original ↗Export Markdown