New AI models
Every day we read hundreds of stories. Only the most important ones are here — short, in our own words, with a link to the original source.
- Artificial intelligence · New AI models
OpenAI launches wet-lab science benchmark
OpenAI introduces BioBasicsBench, a 100-task benchmark testing AI agents on raw experimental data. It measures how well such agents handle everyday wet-lab science tasks.
Full story at openai.com ↗ - Artificial intelligence · New AI models
Egyptian, Nigerian English added to ASR leaderboard
The Open ASR Leaderboard now includes Egyptian and Nigerian English dialects. This allows better comparison of speech recognition models for these varieties.
Full story at huggingface.co ↗ - Artificial intelligence · Programming & dev tools · New AI models
Coding agents struggle on real-world tasks
Alibaba's new Co-Design Bench tests agents on real user requests that already tripped them up. Even the top performer, GPT-6 Astra, fully solves only 26.4 percent, while older benchmarks show above 99 percent.
Full story at The Decoder ↗ - Artificial intelligence · New AI models · Programming & dev tools
Karpathy builds a curated video site with Claude
Andrej Karpathy used Claude's Opus 5.5 to filter 43 million Hacker News comments into 6,454 curated videos. He says the result beats YouTube's own recommendations.
Full story at The Decoder ↗ - Artificial intelligence · New AI models · Enterprise software
Mistral presents Le Chonk and Mistral Studio for enterprises
Mistral AI promotes Mistral Studio together with Le Chonk. Companies can build and deploy agents, workflows and apps faster, with full control over their models, data and infrastructure.
Full story at Mistral AI ↗
An AI writes the short texts from the source's stories.
Free by e-mail
Choose your topics. You only get the most important news on them.