DeepMind Launches Lyria 3.5 in Google Flow Music
DeepMind announced Lyria 3.5 on July 29, 2026, with improvements in musicality, lyrics, vocals, and creative control.
Model Release
432 articles
New AI models as they ship — large language models, multimodal systems, and open-weight releases. Capabilities, context windows, pricing, and availability, summarized so you can compare what is actually new.
LATEST NEWS
-
-
OpenAI Launches ChatGPT for Academic Researchers
OpenAI is providing free access to its frontier models to 100,000 researchers at selected academic institutions by 2027.
-
Tether AI Releases VisionPsy-Nano, Compact On-Device VLMs
Tether AI Research launches VisionPsy-Nano, a family of ~460M-parameter vision-language models (VLMs) that outperform peers on 16 of 17 benchmarks with a normalized score of 62.3.
-
OpenAI Releases GPT Transcribe With Faster Processing
OpenAI's GPT Transcribe processes audio 34 times faster than real time, achieving a 3.31% word error rate, a 0.7 percentage point improvement over its predecessor.
-
Pangram Raises $9M to Combat AI Content Flood
New York-based AI detection startup Pangram raised $9 million to tackle the surge in AI-generated content online, launching its fourth-generation text detection model with over 99% accuracy.
-
1X Unveils Robotic Hands With Dexterous Movements
1X demonstrated robotic hands in July, sparking debate over their appeal and raising questions about AI's role in daily tasks.
-
Hugging Face Releases FeyNoBg, State-of-the-Art Background Removal Model
Hugging Face introduces FeyNoBg, a background removal model that achieves the best published S-measure on four benchmarks and closely matches leaders on others.
-
Moonshot AI Unveils Kimi K3 Open Weights and Infrastructure
Moonshot AI released Kimi K3 model weights and infrastructure in July 2026, outperforming Western models like Fable 5 and GPT-5.6 Sol on benchmarks.
-
Microsoft unveils AI security tools claiming outperformance over rivals
Microsoft announced AI tools designed to automate security risk management, scoring 96 on CyberGYM, 12 points higher than Anthropic's Mythos.
-
Microsoft Launches Cybersecurity Model MAI-Cyber-1-Flash
Microsoft's MAI-Cyber-1-Flash scores 96 percent on CyberGym, outperforming Gemini and GPT by 12 points.
-
Meta AI Now Available in Threads Direct Messages
Meta announced on Monday that its AI chatbot is now accessible via direct messages in Threads, expanding private interactions with the assistant.
-
Guardoc Uses Amazon Nova Models to Improve Medical Documentation
Guardoc Health reports a 46% reduction in documentation errors using Amazon Nova models, helping facilities save over $400K annually.