Kakao’s Kanana2 leads Google and Alibaba’s lightweight AI models in safety evaluation
The Kanana2 1.3B and 3B models ranked first in the overall safety evaluation.
Modelle, Agenten, Anwendungen: künstliche Intelligenz, die an echten Problemen arbeitet.
The Kanana2 1.3B and 3B models ranked first in the overall safety evaluation.
Tesla is preparing for public Cybercab operations, which could begin as early as August in Austin, Texas.
Amazon India, Meesho and Flipkart are bringing AI tools to online sellers.
Reddit is testing AI voices that read posts and comments in video form.
MIT researchers made bacteria act as switches for biological circuits.
AMD says AI delivered a 30 percent productivity boost in one year.
Hi3D V3.0 raises AI 3D modeling resolution to 2048³ voxels.
TensorCast gives model tensors one programmable lifecycle across compute, networks and storage.
IntelliCAD 15.0 previews plain-language AI workflows inside drawings.
RealReplicaBench tested AI on 107 real-world e-commerce tasks; the top score among 13 models was just 56.1 points.
The results of the second evaluation of the sovereign AI foundation model project will be released soon.
Using AI increased video-search productivity by 3–35%, but reduced accuracy by up to 32%.
The number of Qwen derivative models on Hugging Face Hub reached 151,448.
Meshy 7 measures how closely a generated 3D model matches its source image.
Cloudflare will KI-Agenten per Wallet kostenpflichtige APIs bezahlen lassen.
- A charged fluoropolymer surface lets a robot detect nearby objects without physical contact.
Maintaining its 700-billion-parameter scale, GLM-5.3 relies primarily on post-training to improve coding and agent capabilities.
- Zhipu released GLM-5.3 hours after its chief scientist teased “sooooooon.”