analiza aiqbench
Google DeepMind's Gemma 4 12B eliminates separate vision and audio encoders, feeding raw modalities directly into a unified decoder-only backbone to reduce latency and parameter overhead. This architecture enables agentic workflows on consumer hardware with as little as 16 GB of RAM while maintaining an open Apache 2.0 license for local deployment. The release bridges the gap between edge-friendly and larger MoE variants, offering a practical path for running multimodal models on laptops and Apple Silicon without external inference stacks.
Google DeepMind Releases Gemma 4 12B: An Encoder-Free Multimodal Model with Native audio that runs on a 16 GB laptop
Gemma 4 12B feeds vision and audio straight into the LLM backbone, running locally under an Apache 2.0 license. The post Google DeepMind Releases Gemma 4 12B: An Encoder-Free Multimodal Model with Native audio that runs on a 16 GB laptop appeared first on MarkTechPost.
Google DeepMind just released Gemma 4 12B, a dense multimodal model that strips out traditional encoders entirely. Vision and audio flow straight into the LLM backbone. The result is a model that runs agentic workflows on a consumer laptop with 16 GB of RAM. It ships under the Apache 2.0 license. Gemma 4 12B is a 12-billion-parameter decoder-only transformer. It handles text, images, audio, and video natively. There are no separate vision or audio encoders. The decoder uses the same structure…
Extras afișat conform termenilor RSS ai publicației · 81 cuvinte · textul integral și figurile sunt pe site-ul publicației
Subiecte conexe
Google Gemma 4 12B: Architecture, Benchmarks, Access, and Hands-on Guide for Developers
Analytics Vidhya · 5 iun. 2026
[AINews] Quasi-Riemann-Hypothesis: OpenAI publishes 722 math papers solving 90 of the top 500 open math problems; “the most significant moment” in >100 years of mathematics
Latent Space · 7 oct. 2026
[AINews] Modal’s $1B ARR, $16B Series D makes it NYC’s newest AI Infra Decacorn, 3x usage and 3x valuation in 5 months, 15x in 1 yr
Latent Space · 15 oct. 2026
Introducing Gemma 4 12B: a unified, encoder-free multimodal model
Google DeepMind · 9 iun. 2026