Meta launches new multimodal Llama-4 AI models

Meta has announced the release of its latest Llama-4 series models, which introduce a multimodal architecture built from scratch to handle both image and text inputs. The new models boast massive context windows of up to 10 million tokens, positioning them as formidable competitors to industry rivals like GPT-4 and Gemini 2.0. These releases highlight Meta’s strategic push into advanced AI capabilities while also sparking discussion about regional availability and market impact.

11952 simonwillison.net / Quoting Andriy Burkov

GPT-4.5 and Llama 4 show that bigger isn’t better without reinforcement learning aimed at reasoning. RL works only when rewards can be assigned, leaving non-target domains open for tailored, business-specific datasets.

11585 the-decoder.com / Meta releases first multimodal Llama-4 models, leaves EU out in the cold

Meta has released the first two models in its Llama 4 series, marking the company’s initial deployment of a multimodal architecture built from the ground up. The article Meta releases first multimodal Llama-4 models, leaves EU out in the cold appeared first on THE DECODER.

11586 winbuzzer.com / Meta Unveils New Llama 4 AI Models With Massive Context Windows up to 10 Million Tokens

Meta has launched Llama 4 Scout and Maverick with image-text fusion, new architecture, and early results rivaling GPT-4o and Gemini 2.0 in benchmarks. The post Meta Unveils New Llama 4 AI Models With Massive Context Windows up to 10 Million Tokens appeared first on WinBuzzer.

3 stories from sources in 58.4 hour(s) #ai #machine-learning #meta #openai

Related Tags

Meta launches new multimodal Llama-4 AI models

Related Tags

Artificial Intelligence

Machine Learning

Meta

OpenAI