Beat
A new series of open-source models claims to improve agentic coding tasks through a self-scaffolding approach, according to initial user impressions.
On a report by Simon Willison
Chinese and Japanese firms launch AI tools as US export controls limit access to Anthropic’s models.
On a report by TechCrunch AI
High training costs and compressed margins challenge AI labs' financial sustainability.
On a report by Simon Willison
AIpressr is the press wire for the AI era. Optimized to be cited by ChatGPT, Perplexity, Gemini, and the desks that follow them.
Submit a release →
Simon Willison reports that 6,000 attempts failed to breach his AI assistant, highlighting improved defenses against prompt injection.
On a report by Simon Willison
OpenAI introduces three new GPT-5.6 models, with plans for broader release after government-approved preview.
On a report by Simon Willison
General Intuition secures funding to develop AI models trained on video game data for real-world applications.
On a report by TechCrunch AI
Hybrid language models excel at predicting meaningful tokens but struggle with repetitive elements, according to Hugging Face Blog.
On a report by Hugging Face Blog
PaddleOCR's latest model family scales from 1.5M to 34.5M parameters, supporting 50 languages and improved accuracy.
On a report by Hugging Face Blog
Sakana AI's Fugu Ultra, a multi-agent model, is now accessible via AI Gateway, offering competitive reasoning capabilities.
On a report by Vercel
Subquadratic alleges its sparse-attention LLM outperforms transformers in speed and cost.
On a report by MIT Tech Review AI
Z.ai's GLM-5.2 claims top spot on benchmarks but reportedly consumes more tokens than rivals.
On a report by Simon Willison
MolmoMotion predicts object trajectories in 3D space using video frames and action descriptions.
On a report by Hugging Face Blog
Hugging Face's latest model claims improved performance on long-horizon coding tasks with a 1M-token context.
On a report by Hugging Face Blog
Georgi Gerganov highlights Qwen3.6-27B's utility in daily coding tasks, though usage remains limited by PR reviews.
On a report by Simon Willison
Vercel's latest AI model update increases context window to 1M tokens for long-horizon tasks.
On a report by Vercel
Personality conflicts reportedly disrupted Anthropic's AI models, raising questions about jailbreak vulnerabilities.
On a report by Simon Willison
