AI news
1. Meta Releases Llama 5: Open--Weights Model Surpasses Proprietary Reasoning Benchmarks
Meta has officially released Llama 5, marking a significant milestone as the first open--weights model to outperform proprietary systems in complex reasoning and mathematical synthesis. The model features a "Sparse--Attention Mixture" (SAM) architecture, which reduces computational overhead by 60% compared to Llama 4. Meta's leadership emphasized that this release aims to "democratize frontier--level intelligence" for global developers. The company confirmed that the Llama 5 Multimodal Suite will begin its rollout to Ray--Ban Meta smart glasses on November 15, 2026.
| Metric | Llama 5 Performance |
|---|---|
| Reasoning (MMLU--Pro) | 89.4% |
| Inference Efficiency | +60% vs Llama 4 |
| Architecture | Sparse--Attention Mixture |
Source: Meta AI Newsroom