Neura News

AI News

News reporting focused on AI and machine learning, covering the companies behind these technologies, their real-world applications, and the ethical concerns they raise. This includes areas like generative AI (large language models, text-to-image and video), speech tech, and predictive analytics.

Latest News

28 articles
AI Models

Z.ai Releases GLM-5.3 With Big Coding and Cyber Gains, Open Weights Coming in Two Weeks

Z.ai released GLM-5.3 on August 14, 2026, with major improvements in coding and cybersecurity, achieved through scaled-up post-training rather than architecture changes. The model shows significant gains on long-horizon benchmarks like Terminal-Bench 3.0 and DeepSWE, and unexpectedly strong cyber exploitation capabilities. Open weights will be released in about two weeks after safety evaluation.

Aug 147 minNeura News
AI Tools

NVIDIA Launches Magpie Multilingual TTS, an Open-Weight Voice Model for 12 Languages

NVIDIA has released Magpie Multilingual TTS, an open-weights text-to-speech model supporting 12 languages, including new additions like Arabic, Korean, and Brazilian Portuguese. The 364M-parameter model is designed for low-latency voice agents, offering developers full control over deployment, from private servers to air-gapped environments. Benchmarks show sub-200ms end-to-end latency on B200 GPUs, with architectural improvements like frame stacking and local transformers enhancing speed. The model is available on Hugging Face under the NVIDIA Open Model License, paired with NIM microservices and NeMo for fine-tuning.

Aug 107 minNeura News
Research

World Rehearsal Lets LLM Agents Learn Without an Environment

A new training method called EnvACE lets large language model agents practice inside their own heads instead of relying on external environments. The approach, described in a paper submitted to arXiv on 6 Aug 2026, replaces real-world interaction during training with what the authors call world rehearsal. The result is a policy that internalizes how actions change the world, and it performs strongly across four benchmarks.

Aug 74 minNeura News
AI Models

Mistral's Shieldstral Packs Big-Model Safety Into a 3-Billion-Parameter Package

Mistral has released Shieldstral, a 3-billion-parameter open-weight safety classifier that matches the performance of much larger models on text safety benchmarks and sets a new high score for joint text and image classification. The model uses a runtime-definable yes/no question format, allowing operators to write custom screening rules in plain language. Shieldstral is designed to be fast and cost-effective, making it practical for real-time content moderation.

Aug 55 minNeura News
AI Models

Thinking Machines Releases Inkling Small, a Compact Reasoning Model That Outperforms Its Bigger Sibling on Key Tests

Thinking Machines released Inkling Small, a compact open-weights reasoning model that scores nearly as high as its larger sibling on the Intelligence Index (40 vs. 41) and outperforms it on specific reasoning benchmarks like Humanity's Last Exam and GPQA Diamond. The model is highly token-efficient, supports multimodal inputs, and is available under Apache 2.0 with browser-based fine-tuning.

Jul 313 minNeura News
AI Models

Microsoft AI shifts to small specialist models, cuts costs and challenges frontier giants

Microsoft AI is pivoting from large general-purpose models to small specialist models, prioritizing token efficiency and cost reduction. CEO Mustafa Suleyman announced the strategy on July 30, 2026, with early results showing MAI-Cyber-1-Flash outperforming Anthropic's Mythos on cybersecurity benchmarks at half the cost. The shift includes an orchestration system called MDASH that routes tasks to cheaper specialists while reserving frontier models for complex problems, challenging the dominance of monolithic AI models.

Jul 302 minNeura News
Funding

Prentis AI Lab Co-Founded by Reid Hoffman, Marc Pincus Seeks $100M

Prentis, a new AI research lab co-founded by Ritankar Das, Reid Hoffman, and Marc Pincus, is in talks to raise $100 million at a $1 billion valuation. The startup focuses on computer use models that automate office workflows. It has already signed contracts worth up to $50 million with several customers and claims its Hive-32B model outperforms rivals like OpenAI's GPT-5.4 and Anthropic's Claude Opus 4.6 on key benchmarks.

Jul 244 minNeura News