Nvidia has unveiled Lyra 2.0, a new AI model designed to generate explorable 3D environments in real time. The system enables users to move freely through AI-generated worlds while maintaining spatial consistency, addressing a major limitation in existing generative video and 3D models.

Lyra 2.0 builds on recent advances in video-based scene generation, where AI models create camera-controlled walkthroughs and convert them into 3D environments. The model allows users to navigate dynamically through these spaces, effectively rendering new parts of the world as they explore. This approach supports real-time interaction and opens the door to applications in simulation, gaming, and robotics.

Today, we released Lyra 2.0, a framework for generating persistent, explorable 3D worlds at scale, from NVIDIA Research.

Generating large-scale, complex environments is difficult for AI models. Current models often “forget” what spaces look like and lose track of movement over… pic.twitter.com/l6oTNMl5mV

— NVIDIA AI Developer (@NVIDIAAIDev) April 15, 2026

A key differentiator is Lyra 2.0’s ability to maintain consistent geometry over long sequences. Competing systems often struggle with tracking motion over time, leading to visual artifacts such as shifting objects, blurring, or inconsistent scene reconstruction. Nvidia’s model is designed to overcome these issues, enabling stable navigation across complex environments without degradation.

Fixing Drift and Inconsistency in 3D Generation

Lyra 2.0 tackles two core technical challenges: motion drift and spatial inconsistency. In many generative systems, small errors accumulate as the model produces frames over time, eventually distorting the scene. At the same time, previously generated areas may be forgotten, causing the model to recreate them inaccurately when revisited.

To address this, Lyra 2.0 maintains a form of spatial memory by storing per-frame geometry. This allows the model to reference previously seen areas and preserve structural consistency. It also uses a training approach that exposes the system to its own imperfect outputs, helping it learn to correct errors instead of amplifying them.

The result is a system capable of generating longer, more coherent 3D sequences, even when users move in different directions or revisit earlier parts of the environment.

Real-Time Exploration and Simulation Potential

Lyra 2.0 includes an interactive interface that lets users explore generated environments freely, rather than following a fixed path. As users move, the system continuously expands the world, generating new regions while maintaining alignment with previously created structures.

The generated environments can also be exported into simulation tools such as NVIDIA Isaac Sim, making them suitable for robotics training and testing. This could reduce the time and cost required to build large-scale simulation environments.

The release highlights Nvidia’s broader push into generative AI for spatial computing. By combining video generation with 3D reconstruction and real-time interaction, Lyra 2.0 moves closer to enabling fully AI-generated virtual worlds that can be explored, manipulated, and deployed across industries.

OpenAI has introduced GPT-Rosalind, a new domain-specific AI model designed to support research in biology, drug discovery, and translational medicine. The model is being released as a research preview through a controlled access program, reflecting both its advanced capabilities and the sensitivity of its potential applications.

GPT-Rosalind is built to address one of the most complex challenges in life sciences: the fragmented and time-intensive workflows that underpin early-stage discovery. Developing a new drug can take 10 to 15 years, with early research decisions having compounding effects on downstream outcomes. The model is designed to help scientists navigate large volumes of literature, datasets, and experimental variables more efficiently, while also generating and testing new hypotheses.

The system is available through ChatGPT, Codex, and the API for qualified enterprise users. OpenAI is also launching a Life Sciences research plugin for Codex, enabling integration with more than 50 scientific databases and tools. Early collaborators include major pharmaceutical and research organizations such as Amgen, Moderna, Allen Institute, and Thermo Fisher Scientific.

Built for Complex Scientific Workflows

Unlike general-purpose AI models, GPT-Rosalind is optimized for reasoning across specialized domains including chemistry, genomics, protein engineering, and disease biology. It is designed to assist with multi-step research tasks such as literature review, experimental planning, sequence analysis, and data interpretation.

OpenAI reports that the model shows improved performance on benchmarks related to biochemical reasoning, including protein structure analysis, phylogenetics, and experimental design. It also demonstrates stronger ability to use external tools and databases within complex workflows, a critical requirement for real-world scientific research.

In industry evaluations, GPT-Rosalind achieved leading results on bioinformatics benchmarks such as BixBench and outperformed earlier models on several tasks in LABBench2, including molecular cloning design. In collaboration with Dyno Therapeutics, the model also ranked above most human experts on certain RNA prediction tasks.

Controlled Access and Research Integration

Given the potential risks associated with advanced biological research tools, OpenAI is deploying GPT-Rosalind through a “trusted access” model. Organizations must meet criteria related to legitimate scientific use, governance, and security controls before gaining access. The rollout initially focuses on enterprise users in the United States.

The accompanying Life Sciences plugin provides an orchestration layer for scientific workflows, connecting researchers to public datasets, literature sources, and domain-specific tools. This allows the model to move beyond static responses and actively support research processes such as protein structure lookup, sequence search, and dataset discovery.

OpenAI said the system was developed with enhanced security measures and is intended for use in controlled research environments. During the preview phase, usage will not consume standard API credits, though safeguards are in place to prevent misuse.

The release marks the first step in a broader effort to build AI systems tailored to scientific discovery. OpenAI says future iterations will expand the model’s capabilities for long-horizon, tool-intensive workflows, with ongoing collaborations across academia, biotech, and national laboratories aimed at advancing areas such as protein and catalyst design.

A research team from Nanyang Technological University has developed a new AI-powered biochip capable of rapidly detecting microRNAs, tiny genetic markers linked to diseases including cancer and cardiovascular conditions. The system, described in the journal Advanced Materials, combines nanophotonic sensing with automated image analysis to significantly reduce diagnostic time.

The platform can analyze a small blood sample and detect multiple microRNA biomarkers in about 20 minutes, compared to several hours required by traditional methods such as PCR (polymerase chain reaction). Researchers say the system achieves high sensitivity and accuracy, detecting extremely low concentrations of microRNAs, even down to a few molecules.

The work was led by Associate Professor Chen Yu-Cheng, who said the goal is to create a scalable diagnostic platform capable of screening multiple disease markers quickly and accurately. Initial testing focused on microRNAs linked to non-small cell lung cancer, demonstrating the system’s ability to identify multiple targets simultaneously without complex sample preparation.

How the Technology Works

At the core of the system is a nanophotonic chip embedded with nanocavities, microscopic structures that enhance fluorescent signals when microRNAs bind to specific probes. These cavities amplify weak signals, making it possible to detect even single molecules.

The chip is paired with an AI imaging system that captures and analyzes thousands of signals in a single snapshot. Using a deep learning model based on Mask R-CNN, the system automatically identifies and classifies microRNA signals, removing the need for manual counting and reducing the risk of human error.

Unlike conventional approaches, which often require amplification or labeled probes, the NTU platform directly measures microRNAs in liquid samples. The researchers report accuracy levels exceeding 99 percent across test scenarios, including experiments using both cancer cell extracts and synthetic samples.

Toward Faster, Scalable Diagnostics

The team has also built a compact prototype that includes a camera and a mobile application for real-time analysis. This setup could support point-of-care testing, where results are generated quickly without the need for specialized laboratory infrastructure.

Researchers believe the platform could eventually be adapted for large-scale screening, potentially analyzing hundreds or thousands of biomarkers from blood, saliva, or urine samples. This could open the door to earlier disease detection, better monitoring of treatment response, and more personalized healthcare.

Independent experts note that microRNAs have long been considered promising biomarkers but have been difficult to measure reliably due to their small size and similarity. A system that can accurately detect multiple microRNAs could improve clinical decision-making, particularly in oncology and chronic disease management.

The project is supported by Singapore’s research funding programs, and the team has filed a technology disclosure through NTU’s commercialization arm. Future work will focus on clinical validation and scaling the platform for broader use in healthcare and pharmaceutical research.