In a day marked by advancements in AI models and applications, Contrastive-LM’s introduction of CLM-8B stands out, while Meta continues to enhance its AI ecosystem with Muse.
Contrastive-LM Releases CLM-8B: An Open System One Model That Scores Agent Actions Up to 9× Faster Than Jev
Contrastive-LM has launched CLM-8B, an innovative System One model designed to evaluate candidate actions more efficiently than existing models. By incorporating two projection heads into a static Qwen3-8B encoder and utilizing a contrastive InfoNCE objective, CLM-8B demonstrates performance up to nine times faster than TypeSafe’s Jev in zero-shot evaluations. With fine-tuning, it achieves an impressive verification accuracy of 81.6%.
Everything New Coming to Meta’s AI Agent Muse
During Meta’s annual Connect event, CEO Mark Zuckerberg emphasized the company’s commitment to its AI agent Muse, signaling a major push in AI integration across its platforms. Notably, Muse will be featured in Meta’s new AI glasses, showcasing the company’s ambition to create a seamless AI experience for users.
A Coding Guide to TypeSafe AI Jev: Typed Decisions, Calibrated Confidence, and Speculative Fan-Out with a System One Model
A comprehensive tutorial has been released for TypeSafe AI’s Jev, a System One model that focuses on structured decision-making rather than text generation. The guide covers installation of the Python SDK, various question types, and advanced features like speculative fan-out and confidence-gated routing, providing developers with the necessary tools to implement this model effectively.
Anthropic Says Its Biology Lab Has Already Found Something Big
Anthropic has announced significant findings from its biology lab, although it has clarified that its AI model Claude has not been given free rein in these investigations. The lab continues to prioritize human oversight, ensuring that AI remains a supportive tool rather than a replacement in critical research areas.
Google Releases Gemini 3.8 Flash TTS and Flash-Lite TTS With Prompt-Based Voice Design
Google has introduced two new text-to-speech models, Gemini 3.8 Flash TTS and Flash-Lite TTS, which leverage natural language prompts to create voices in over 100 languages. Flash TTS has achieved the highest score on Hume AI’s Voice Design Benchmark, highlighting its advanced capabilities, while Flash-Lite TTS is optimized for high-volume dubbing and voice application use cases.
Enveda Secures $311M to Bring More Nature-Derived AI Drugs into Clinical Trials
Enveda has raised $311 million to advance its AI-driven drug development initiatives, particularly focusing on nature-derived compounds. The funding round values the biotech firm at $2 billion and will support clinical trials for new treatments targeting skin conditions and weight management following the cessation of GLP-1 medications.
NVIDIA Releases Nemotron 3 Diarization: A 100M-Parameter Open-Weight Model That Tracks 8 Speakers in Real Time
NVIDIA has unveiled Nemotron 3 Diarization, a cutting-edge model designed for speaker diarization that can identify and track up to eight speakers in real time. This 100 million-parameter model is capable of handling overlapping voices and is available on Hugging Face, making it a versatile tool for both offline and real-time applications.
YouTube Promises Custom Feeds and a Lot More AI Later This Year
YouTube has announced plans to enhance user experience with custom feeds and increased integration of AI technologies throughout the platform. The upcoming changes are expected to significantly alter content delivery, with a strong emphasis on AI-driven recommendations and enhanced livestream capabilities.
ChatGPT Mobile App Gets Voice-Based Agentic Features
The ChatGPT mobile app has introduced new voice-based features for Pro and Plus users, enabling them to perform agentic tasks directly from their phones. This update enhances the app’s usability, allowing users to leverage AI capabilities more effectively in their daily activities.
Compiled automatically by the Tech AI Newsdesk from public AI-news sources and summarised in our own words.
