105 articles
ZTE's EmbodiedBrain model introduces a unified vision-language-action system designed to improve planning and execution in physical environments, demonstrating state-of-the-art performance across complex embodied AI tasks.
AR3D-R1 marks a breakthrough as the first reinforcement learning-enhanced text-to-3D generation model, arriving with MME-3DR, a dedicated benchmark for testing RL methods in 3D content creation.
Nomos 1, a 30-billion-parameter open-source model, scored 87/120 on the 2025 Putnam Exam, placing it near the top tier of human competitors and demonstrating major progress in accessible AI mathematical reasoning.
DePaola-5 holographic memory model maintains stable retrieval past 500 stored items while classic HRR fails after 125, representing a breakthrough in scalable AI memory systems.
EngineAI released real-world footage showing CEO Zhao Tongyang taking a direct hit from the company's T800 humanoid robot. The demonstration, verified to contain no CGI or AI effects, ignited widespread debate about humanoid robot capabilities and safety standards.
Grok 4.20 has jumped to the top of AI trading rankings, beating every major competitor and racking up $16,159 in total equity. The model now holds three of the top five spots in current standings.
Kling O1's latest demo showcases frame-by-frame replication of complex dance movements, demonstrating how AI is reaching new levels of precision in motion capture and animation.
Google's Titans architecture scored 70% accuracy at ten million tokens on the BABILong benchmark, significantly outperforming leading AI models across extreme sequence lengths.
New analysis suggests that neural networks share low dimensional weight subspaces, but critics argue that a sample of only 500 models is too small to support broad conclusions.
Massive studies across three countries reveal conversational AI can genuinely shift political views, with bigger models packing more persuasive punch. The link between model scale, information depth, and lasting influence is undeniable.
Grok 4.1 debuts with an industry-leading 2-million-token context window, outpacing all major AI competitors and enabling entire codebases and lengthy documents to process in one session.
Arcee released new Trinity Mini and Trinity Nano models built on a mixture-of-experts architecture, featuring efficient active-parameter configurations. Early benchmark results show competitive performance across multiple reasoning and evaluation tasks.
A new AI method called Chain-of-Visual-Thought introduces continuous visual tokens to improve reasoning in vision-language models. Recent tests show accuracy gains of 3–16 percent across major benchmarks.
A new AI framework called REASONEDIT introduces reasoning-enhanced image editing through multimodal thinking. The system interprets abstract instructions and refines visual outputs through multiple generation stages.
Odyssey launched Odyssey-2, an AI video model that generates personalized visual streams from text prompts in real time, removing the typical delays found in current AI video tools.
A worldwide test that covered eight elections showed that five large AI systems lean toward centre left policies - Grok-4 alone matched the real results voters produced.
BrainBridge introduces a futuristic concept for AI-powered head and face transplants using advanced robotics. The blueprint pushes boundaries between cutting-edge medicine and science fiction, raising questions about what's medically possible.
Kimi's strong benchmark performance against major AI models has reignited discussions about its valuation, which sits 4-5 times below competing labs despite delivering competitive results.
Sharp Robotics revealed its SharpaWave robotic hand – a life-size system packing 22 degrees of freedom and over 1,000 tactile sensors in each fingertip. Fresh demos show a humanoid robot cooking full meals, cracking eggs and wielding kitchen tools with startling precision.
A new benchmark compares political orientations of frontier AI models across global elections, revealing consistent ideological patterns. Most systems cluster in the libertarian-left quadrant.
We use cookies to improve your experience on our site and to show you relevant advertising. To find our more, read our privacy policy and cookie policy
Usman Salis
Victoria Bazir
Saad Ullah
Eseandre Mordi
Marina Lyubimova
Artem Voloskovets
Alex Dudov
Peter Smith
