AI Model Releases & Benchmarks
New AI models including world and generative media models: what's new, benchmarks and access.
- OpenAI's LifeVal Tests AI on Everyday Errands, and GPT-6 Astra Scores 80%OpenAI's new LifeVal evaluation scores AI on chores like school searches, meal plans and grocery orders, and the company says GPT-6 Astra reached 80% overall, up from 49.3% for GPT-5.2.
- New GDPx Benchmark Puts Claude Opus 5.5 First, but the Top Score Is Just 12SemiAnalysis's GDPx tests 13 frontier models on 206 expert tasks across 11 domains. Claude Opus 5.5 leads with 12.0, safety filters reshuffle the order, and most real work still fails. Here is what the results mean.
- Meituan's LongCat-2.5 Takes On Claude Opus 5 With 1.6 Trillion ParametersMeituan's new LongCat-2.5 is a 1.6-trillion-parameter multimodal model built for long coding-agent tasks and trained on Chinese chips. Here is how its benchmarks compare with Claude Opus 5 and GPT-5.6-Sol.
- Nvidia's Nemotron 3.5 Transcribe: Open Model Covers 40 Languages and 8 SpeakersNvidia's open-weight Nemotron 3.5 Transcribe, released October 28, transcribes 40 languages, translates 31 to or from English and labels up to eight speakers in recordings up to an hour long.
- Google's Gemini 4 Carbon Tops AI Rankings, but Cyber Defenders Get It FirstGoogle DeepMind's Gemini 4 Carbon takes first place on the Artificial Analysis Intelligence Index with a score of 59 and leads three security benchmarks, but only vetted cyber defenders can use it for now. Here's what it does, what it costs and when you might get access.
- Google's Genie 4 Adds Sound, Speech and 2-Minute AI Worlds After Game Stock RoutGoogle DeepMind's Genie 4 world model, unveiled October 27, adds native audio, interactive dialogue and up to two minutes of consistent real-time worlds, double Genie 3, with Google AI Ultra access coming in the next few weeks.