AI-curated high-value content from 50+ sources
OpenAI has unveiled GPT-5, its most capable model to date, featuring native multimodal reasoning across text, images, audio, and video. The model supports a 2 million token context window and demonstrates significant improvements in mathematical reasoning and code generation. Enterprise API pricing starts at $0.03 per 1K tokens.
Anthropic has closed a $5 billion funding round led by Lightspeed Venture Partners, valuing the company at $60 billion. The funds will be used to expand compute infrastructure and accelerate Claude's development. The company also announced plans to open a European headquarters in London.
DeepMind's latest AlphaFold iteration achieves a 50% improvement in predicting how proteins interact with other molecules, including DNA, RNA, and small molecules. The breakthrough could accelerate drug discovery timelines by years. The model is available through the AlphaFold Server for non-commercial research.
Meta has open-sourced Llama 4, a 400B parameter model with 256K context window support. The model outperforms GPT-4 on several benchmarks while being available for free commercial use. Meta also introduced new safety evaluation tools and a responsible use license.
The European Union's AI Act has entered its first enforcement phase, requiring foundation model providers with over 10^25 FLOPs to comply with transparency obligations. Companies must publish detailed technical documentation and training data summaries. Non-compliance carries fines up to 3% of global annual turnover.
Hugging Face has released Inference API v2, promising up to 10x speedup for popular open-source models through optimized kernel fusion and dynamic batching. The new API supports streaming responses and function calling for models like Llama, Mistral, and Qwen.
Researchers at Microsoft have published a new prompt optimization method called 'AutoPrompt v3' that automatically refines prompts for specific tasks, achieving 15-30% accuracy improvements on benchmark datasets without manual tuning. The technique uses gradient-based optimization on a small validation set.
In a widely-shared blog post, Simon Willison argues that the most impactful AI developments are happening in small, efficient models rather than frontier-scale systems. He highlights how 7B parameter models can now match GPT-3.5 performance on many tasks, enabling local deployment and privacy-preserving applications.