OpenAI has launched a public hub for sharing AI safety evaluations. The portal reveals performance on hallucination, jailbreak, and harmful output tests, updated regularly to boost transparency.
– Mistral AI’s Pixtral models are 60x more prone to generating child exploitation content and 18–40x more likely to generate chemical/biological threats than competing AI models.
– **Slug:** mistral-ai
AWS is developing Kiro, a sophisticated AI coding tool aimed at transforming software development by integrating real-time code generation, multimodal interfaces, and AI agent collaboration.
Hugging Face’s Open Computer Agent runs in a cloud-based Linux VM, executing GUI tasks via Python and vision models. It’s open-source, secure, and free to use.
Gemini 2.5 Pro I/O edition raises the bar in AI with top coding scores, long-context reasoning, and multimodal integration across web and video tasks.
Nova Premier is Amazon’s most capable AI model, designed for enterprise-scale deployment with multimodal input, model distillation, and multi-agent coordination. It excels in cost and orchestration, even if it trails in STEM benchmarks.
Alibaba’s Qwen3 AI models offer hybrid reasoning, outperforming OpenAI and Google in key benchmarks. Open-source, multilingual, and cloud-ready.
Meta Launches Standalone AI App powered by Llama 4, introducing voice features, smart glasses sync, and a social Discover feed.
Officially announced on April 27, DeepSeek-R2 introduces a new AI model with robust multilingual, coding, and multimodal strengths—reinforced by the math-specialist Prover-V2.
