NVIDIA Brings DLSS 4.5 Controls to GeForce NOW Cloud Gaming
NVIDIA is giving cloud streamers finer control over DLSS settings and previewing a faster DLSS 5 model.
AI & ML
Frontier models, research results, agents, and developer tooling.
DeepMind Opens Nano Banana 2 Lite and Gemini Omni Flash to Developers
Google DeepMind is inviting developers to start building with two newly available models, Nano Banana 2 Lite and Gemini Omni Flash.
OpenAI Releases Brazilian Portuguese Educator Guide, Study Mode
The release pairs a localized guide for teachers with a Study Mode feature designed to encourage step-by-step questioning.
Meta’s Muse Spark Open Brings Open-Weight Multimodal MoE to Hugging Face
Hugging Face highlights Meta’s Muse Spark Open, an open-weight multimodal mixture-of-experts model aimed at advancing accessible frontier AI.
FLUX.3 Image Klein Arrives in Hugging Face Diffusers
Hugging Face announces the integration of FLUX.3 Image Klein into its Diffusers library.
Apertus 1.5 Release Notes Published on Hugging Face
Swiss AI has posted release notes for Apertus 1.5 on Hugging Face, though the supplied source contains no technical details.
Granite Switch 4.1 Packs Chat, Decisions, and Speech into One Model
IBM's Granite Switch 4.1 combines language, decision-making, and speech recognition in a single efficient endpoint.
OpenAI offers free GPT-Rosalind access to academic labs
A new academic plan gives eligible life-sciences labs up to 30 free seats for OpenAI's dedicated biology model and workbench for a year.
OpenAI Traces 18-Year-Old Bug Through Core Dump Analysis
OpenAI's large-scale core dump analysis shows how rare infrastructure crashes can be traced to both hardware faults and decades-old software bugs.
OpenAI Previews GPT-5.6 Sol With Coding, Science, Safety Upgrades
OpenAI's next-generation model, GPT-5.6 Sol, promises stronger performance in coding, science, and cybersecurity while debuting the company's most advanced safety stack.
DeepMind and Singapore build on-device AI to flag scams
Google DeepMind and GovTech Singapore fine-tuned a privacy-preserving on-device model to detect manipulative conversations in real time.
Google opens Deep Review AI to verify technical documents
Google's agentic review tool, piloted at top CS conferences, is now publicly available for grounded verification of complex documents.
SentinelOne Uses OpenAI Daybreak to Cut Malware Analysis From Weeks to Hours
OpenAI's Daybreak model lets SentinelOne reverse-engineer malware in about eight hours instead of four to six weeks, enabling broader threat analysis.
Block uses OpenAI Daybreak to compress security response times
The payments company says AI-driven triage and investigation have cut tasks that took weeks down to hours, with humans still validating every result.
Leash Bio Turns DEL Assay Data Into AI Agent Training Ground
A biotech lab's hard-won DNA-encoded library data is now powering verifiable scientific reasoning agents, with an open dataset and preprint released as Minos.
Fish Audio's Drama 3 Brings Open-Weight TTS With 35B MoE
A new open-weight text-to-speech model from Fish Audio uses a 35B mixture-of-experts design while keeping active parameters low.
Third Moment Research runs 10,000 AI agents for 36+ hours
OpenAI spotlights a lab whose agents run for days and get better the more they are used.
OpenAI's Amp expands AI agents beyond coding with new models
Amp uses OpenAI's latest models to extend agent capabilities to non-coding tasks.