Term
Llama 3.2
Llama 3.2 is Meta's model generation from September 2024 — the first with vision support (11B/90B) and lightweight edge models (1B/3B) for mobile and on-device use.
Llama 3.2 — explained in more detail
Meta released Llama 3.2 on September 25, 2024 at the Connect conference. The generation comes in four sizes: two pure text models (1B and 3B) for edge and mobile devices, plus two vision models (11B and 90B) that accept images as input. It was Meta’s first official step into multimodality. All models support a context window of up to 128,000 tokens.
Example / Practical use
The vision variants use separately trained image-reasoning adaptors connected to the language core via cross-attention — typical applications: image classification, object detection, OCR, contextual image Q&A. The vision models process high-resolution images up to 1120×1120 pixels. The small text models (1B/3B) suit summarization, instruction following and text rewriting on smartphone and IoT hardware. Available directly from Meta as well as on AWS Bedrock, Oracle OCI, IBM watsonx, Hugging Face and many others.
Entdecke mehr
Three Days with Claude Fable 5 — and Why Opus Stays My Daily Driver
Fable 5 fixed a bug in one shot that Opus failed at twice. I'm still not switching. My honest impressions from the trenches after three days.
GlossarLlama 4 Maverick
Llama 4 Maverick (April 2025) is Meta's large Mixture-of-Experts variant in the Llama 4 lineup — 17B active out of 400B parameters across 128 experts, 1M-token context.
LexikonComparing AI models — who builds what and how to choose
The major model families in 2026 at a glance. Who builds Claude, GPT, Gemini, Llama, Mistral, DeepSeek, Qwen — and which model to pick when.