Meta has launched its latest AI model family, Llama 4, which aims to engage with more controversial topics, such as politics, than previous versions. The new models, Scout and Maverick, reportedly reject less than 2% of contentious prompts, compared to 7% in Llama 3.3. They have been tested for balance in responses, showing a mere 1% in refusing to answer opposing sides of a debate. Llama 4 models are multimodal, enabling them to process various data types, making them among the most advanced AI systems available.
Meta's latest AI model family, Llama 4, is designed to tackle more contentious topics while maintaining a balanced approach between user engagement and controversy.
With Llama 4, the refusal rate for addressing contentious topics dropped from 7% with its predecessor to under 2%, enhancing its responsiveness.
The new Llama 4 models were tested with controversial questions, demonstrating a refusal rate of only 1% when asked about opposing sides.
Meta claims the Llama 4 Scout and Llama 4 Maverick are their most advanced AI systems yet, capable of multimodal data integration.
Collection
[
|
...
]