Alternatives to NVIDIA Alpamayo 2 Super

NVIDIA Alpamayo represents a comprehensive platform of AI models, simulation resources, and datasets aimed at enhancing the evolution of self-driving vehicles equipped with human-like reasoning abilities. At its core lies a suite of Vision-Language-Action (VLA) models that merge visual analysis, language-based logic, and action strategies, empowering vehicles to navigate intricate driving situations and execute decisions incrementally. In contrast to conventional systems that primarily depend on pattern recognition, Alpamayo incorporates chain-of-thought reasoning, enabling autonomous vehicles to comprehend rare or unexpected "long-tail" events while providing explanations for their actions, thereby fostering increased safety and transparency. Furthermore, it seamlessly integrates with NVIDIA’s complete autonomous driving framework, encompassing aspects of training, simulation, and deployment, allowing developers to create sophisticated systems without the need to build foundational infrastructure from the ground up. With these capabilities, Alpamayo not only enhances the functionality of autonomous vehicles but also contributes to the broader goal of making intelligent transportation solutions more accessible.

Mobileye offers a wide range of advanced driver-assistance systems (ADAS), extending to autonomous vehicles for public transportation and goods delivery, as well as consumer automotive applications. By managing the entire process from silicon production to the implementation of self-driving technology internally, Mobileye harnesses multiple efficiencies and synergies, enabling large-scale deployment of autonomous vehicles. Since its inception, the company has focused on developing both hardware and software in-house, which has led to the creation of highly effective technological stacks that excel in cost-performance efficiency. Every product from Mobileye is designed with safety as a core principle, supported by a comprehensive strategy aimed at making this cutting-edge technology accessible to the general public. This commitment to innovation and safety ensures that Mobileye is well-positioned to lead the future of autonomous driving.

Software transforms a vehicle into a smart machine, and the NVIDIA DRIVE™ Software stack serves as an open platform that enables developers to effectively create and implement a wide range of advanced autonomous vehicle applications, such as perception, localization and mapping, planning and control, driver monitoring, and natural language processing. At the core of this software ecosystem lies DRIVE OS, recognized as the first operating system designed for safe accelerated computing. This system incorporates NvMedia for processing sensor inputs, NVIDIA CUDA® libraries to facilitate efficient parallel computing, and NVIDIA TensorRT™ for real-time artificial intelligence inference, alongside numerous tools and modules that provide access to hardware capabilities. The NVIDIA DriveWorks® SDK builds on DRIVE OS, offering essential middleware functions that are critical for the development of autonomous vehicles. These functions include a sensor abstraction layer (SAL) and various sensor plugins, a data recorder, vehicle I/O support, and a framework for deep neural networks (DNN), all of which are vital for enhancing the performance and reliability of autonomous systems. With these powerful resources, developers are better equipped to innovate and push the boundaries of what's possible in automated transportation.

We provide licensing for AI software that spans the entire L2-L4 autonomous driving framework, which includes components like perception, intent modeling, path planning, and vehicle control. Our solutions achieve exceptional accuracy in perception and intent prediction, significantly enhancing the safety of autonomous driving systems. By leveraging unsupervised learning alongside mathematical modeling, we can harness vast datasets for improved performance, bypassing the limitations of supervised learning. These advancements lead to technologies that are remarkably more capital-efficient, resulting in a reduced development cost for our clients. Our offerings include Helm.ai's comprehensive scene vision-based semantic segmentation, integrated with Lidar SLAM outputs from Ouster. We facilitate L2+ autonomous driving capabilities with Helm.ai on highways 280, 92, and 101, which encompasses features such as lane-keeping and adaptive cruise control (ACC) lane changes. Additionally, Helm.ai excels in pedestrian segmentation, utilizing key-point prediction to enhance safety. This includes sophisticated pedestrian segmentation and accurate keypoint detection, even in challenging conditions like rain, where we address corner cases and integrate Lidar-vision fusion for optimal performance. Our full scene semantic segmentation also accounts for various road features, including botts dots and faded lane markings, ensuring reliability across diverse driving environments. Through continuous innovation, we aim to redefine the boundaries of what autonomous driving technology can achieve.

NVIDIA Cosmos serves as a cutting-edge platform tailored for developers, featuring advanced generative World Foundation Models (WFMs), sophisticated video tokenizers, safety protocols, and a streamlined data processing and curation system aimed at enhancing the development of physical AI. This platform empowers developers who are focused on areas such as autonomous vehicles, robotics, and video analytics AI agents to create highly realistic, physics-informed synthetic video data, leveraging an extensive dataset that encompasses 20 million hours of both actual and simulated footage, facilitating the rapid simulation of future scenarios, the training of world models, and the customization of specific behaviors. The platform comprises three primary types of WFMs: Cosmos Predict, which can produce up to 30 seconds of continuous video from various input modalities; Cosmos Transfer, which modifies simulations to work across different environments and lighting conditions for improved domain augmentation; and Cosmos Reason, a vision-language model that implements structured reasoning to analyze spatial-temporal information for effective planning and decision-making. With these capabilities, NVIDIA Cosmos significantly accelerates the innovation cycle in physical AI applications, fostering breakthroughs across various industries.

Waymo, a pioneer in autonomous driving technology, focuses on the development of self-driving vehicles and offers fully driverless transportation services. Initially launched as Google's self-driving car initiative in 2009, it evolved into a standalone subsidiary of Alphabet with the mission of enhancing safety, accessibility, and efficiency in transportation through the use of autonomous technology. Central to its operations is the Waymo Driver, a sophisticated system that integrates artificial intelligence with high-resolution cameras, radar, lidar sensors, and intricate digital maps, enabling vehicles to understand their environment and traverse roads autonomously. The system is designed to constantly evaluate traffic signals, pedestrians, other vehicles, and road conditions to make immediate driving decisions that prioritize safety. Prior to entering a new geographic location, Waymo meticulously maps the area, capturing detailed information about lane markings, signage, and intersections, which is then paired with real-time sensor data to ensure accurate vehicle positioning. This comprehensive approach not only enhances the effectiveness of its technology but also ensures a reliable and secure driving experience for passengers.

Cognata provides comprehensive simulation solutions for the entire product lifecycle aimed at developers of ADAS and autonomous vehicles. Their platform features automatically generated 3D environments along with realistic AI-driven traffic agents, making it ideal for AV simulation. Users benefit from a readily available library of scenarios and an intuitive authoring tool to create countless edge cases for autonomous vehicles. The system allows for seamless closed-loop testing with straightforward integration. It also offers customizable rules and visualization options tailored for autonomous simulation, ensuring that performance is both measured and monitored effectively. The digital twin-grade 3D environments accurately reflect roads, buildings, and infrastructure, down to the finest details such as lane markings, surface materials, and traffic signals. Designed to be globally accessible, the cloud-based architecture is both cost-effective and efficient from the outset. Closed-loop simulation and integration with CI/CD workflows can be achieved with just a few clicks. This flexibility empowers engineers to merge control, fusion, and vehicle models seamlessly with Cognata's comprehensive environment, scenario, and sensor modeling capabilities, enhancing the development process significantly. Furthermore, the platform's user-friendly interface ensures that even those with limited experience can navigate and utilize its powerful features effectively.

Embotech has developed PRODRIVER to address the challenges of motion planning in autonomous or highly automated vehicles. This crucial element resides within the 'decision making' layer of the software architecture for autonomous driving. As a motion planner, PRODRIVER generates either drivable trajectories or direct actuator commands, such as steering, acceleration, and braking, based on the information it gathers from the surrounding environment. It achieves this by continuously predicting scenarios and solving optimization problems in real time. Key inputs for PRODRIVER include data regarding the navigable area, obstacles present, and a defined goal, which might be a specific location or an overarching objective like advancing along a path. The outputs produced can either be directly utilized to steer the vehicle or serve as set-points for the low-level controllers to maintain control. Additionally, the schematic diagram below illustrates how PRODRIVER fits into a typical software stack for autonomous vehicles, showcasing its integral role in ensuring safe and efficient navigation.

The NVIDIA Llama Nemotron family comprises a series of sophisticated language models that are fine-tuned for complex reasoning and a wide array of agentic AI applications. These models shine in areas such as advanced scientific reasoning, complex mathematics, coding, following instructions, and executing tool calls. They are designed for versatility, making them suitable for deployment on various platforms, including data centers and personal computers, and feature the ability to switch reasoning capabilities on or off, which helps to lower inference costs during less demanding tasks. The Llama Nemotron series consists of models specifically designed to meet different deployment requirements. Leveraging the foundation of Llama models and enhanced through NVIDIA's post-training techniques, these models boast a notable accuracy improvement of up to 20% compared to their base counterparts while also achieving inference speeds that can be up to five times faster than other leading open reasoning models. This remarkable efficiency allows for the management of more intricate reasoning challenges, boosts decision-making processes, and significantly lowers operational expenses for businesses. Consequently, the Llama Nemotron models represent a significant advancement in the field of AI, particularly for organizations seeking to integrate cutting-edge reasoning capabilities into their systems.

A combination of sensors, including LiDAR, cameras, and radar, gather data from the vehicle's surroundings. By employing sensor fusion technology, perception algorithms are capable of identifying, locating, measuring the speed, and determining the orientation of various objects on the road in real time. This advanced autonomous perception system is supported by Baidu's extensive big data infrastructure and deep learning capabilities, along with a rich repository of labeled real-world driving data. The robust deep-learning platform, complemented by GPU clusters, enhances processing power. Additionally, the simulation environment enables virtual driving across millions of kilometers each day, leveraging diverse real-world traffic and autonomous driving data. Through this simulation service, partners can access an extensive array of autonomous driving scenarios, allowing for rapid testing, validation, and optimization of models in a manner that prioritizes both safety and efficiency, ultimately fostering advancements in autonomous vehicle technology.

DriveMod represents Cyngn's comprehensive solution for autonomous driving, seamlessly integrating with commonly available sensing and computing equipment to empower industrial vehicles with the ability to understand their environment, make informed decisions, and execute actions. This innovative system is designed to fit effortlessly into your current operations, allowing for straightforward programming of vehicle routes, loops, and missions. Essentially, anything a human driver can accomplish, DriveMod is capable of achieving as well. You can safely equip any commercially available vehicle with autonomous features through a simple retrofit process. The adaptability of DriveMod guarantees that diverse fleets operate efficiently, regardless of the vehicle's make or model. By leveraging advanced AI software alongside top-tier sensors and computing technology, DriveMod delivers performance that surpasses that of human operators. It can identify thousands of objects and evaluate numerous potential paths, efficiently determining the best route in mere fractions of a second, thereby revolutionizing the way vehicles navigate their surroundings. This remarkable capability positions DriveMod as a leading solution in the realm of autonomous vehicle technology.

NVIDIA DRIVE® Map is an advanced mapping platform crafted to support the utmost levels of vehicle autonomy while enhancing safety measures. By merging precise ground truth mapping with the agility and scale of AI-driven fleet-sourced mapping, it achieves remarkable results. The system utilizes four distinct localization layers—camera, lidar, radar, and GNSS—ensuring the necessary redundancy and flexibility for sophisticated AI drivers. With a focus on exceptional accuracy, the ground truth map engine generates DRIVE Maps by integrating a variety of sensors, including cameras, radars, lidars, and differential GNSS/IMU, all captured through NVIDIA DRIVE Hyperion data collection vehicles. It delivers an impressive accuracy of better than 5 cm, particularly in high autonomy scenarios (L3/L4), in environments like highways and urban areas. Designed for rapid operation and global adaptability, DRIVE Map leverages both ground truth and fleet-sourced information, encapsulating the shared knowledge of millions of vehicles on the road. This innovative approach not only enhances mapping precision but also contributes to the evolving landscape of autonomous driving technology.

MORAI presents an innovative digital twin simulation platform designed to expedite the development and evaluation of autonomous vehicles, urban air mobility solutions, and maritime autonomous surface vessels. This platform utilizes high-definition mapping and an advanced physics engine to seamlessly connect real-world applications with simulated testing environments, ensuring all critical components for validating autonomous systems are included, such as those for self-driving cars, drones, and unmanned marine vehicles. It features a comprehensive array of sensor models, which encompass cameras, LiDAR, GPS, radar, and Inertial Measurement Units (IMUs). Users have the capability to create intricate and varied testing scenarios derived from actual data, including those based on logs and edge cases. Furthermore, MORAI's cloud-based simulation framework enables safe, efficient, and scalable testing processes, allowing multiple simulations to operate simultaneously while assessing various scenarios in parallel. This robust infrastructure not only enhances the reliability of testing but also significantly reduces the time and costs associated with the development of autonomous technologies.

Wayve stands out as a pioneering platform for autonomous driving technology, leveraging AI foundation models to fuel the development of future self-driving vehicles with its innovative Embodied AI strategy. The centerpiece of Wayve's advancement is a self-learning “AI driver” that empowers vehicles to interpret, anticipate, and maneuver through intricate real-world scenarios by acquiring knowledge through experience instead of depending on pre-programmed rules or detailed maps. By utilizing primarily camera inputs and deep learning techniques, this system cultivates a versatile driving intelligence capable of adjusting to new roads, urban landscapes, and various vehicle types with minimal need for retraining. Wayve's approach features a mapless and hardware-agnostic framework that allows automobile manufacturers to introduce sophisticated driver assistance and autonomous functions via software updates, accommodating automation levels ranging from L2+ to L4. This innovative design is intended to perpetually learn from both real-world experiences and simulated environments, fostering safe and instinctive driving behavior while enhancing the vehicle's response to unforeseen circumstances. With its focus on adaptability and continuous improvement, Wayve aims to redefine how self-driving technology integrates into everyday transportation.

Aptiv is an international technology firm dedicated to creating safer, more sustainable, and interconnected solutions that pave the way for the future of transportation. The company concentrates on innovating and commercializing autonomous vehicles and systems that facilitate efficient point-to-point transportation through extensive fleets of self-driving cars, particularly in complex urban settings. With skilled teams located worldwide, from Boston to Singapore, Aptiv has emerged as the first organization to launch a commercial autonomous ride-hailing service in Las Vegas. They have successfully completed over 100,000 rides for the public, with an impressive 98% of passengers giving their self-driving experience a perfect 5-out-of-5 star rating. Aptiv is committed to the belief that their mobility innovations can significantly impact the world, and they continue to strive for advancements that enhance the quality of urban transport. By focusing on safety and efficiency, Aptiv aims to redefine how people navigate through cities in the future.

The Nemotron-3 Super is an innovative member of NVIDIA's Nemotron 3 series of open models, specifically crafted to facilitate sophisticated agentic AI systems that can effectively reason, plan, and carry out multi-step workflows in intricate environments. This model features a unique hybrid Mamba-Transformer Mixture-of-Experts architecture that merges the streamlined efficiency of Mamba layers with the contextual depth provided by transformer attention mechanisms, which allows it to adeptly manage extended sequences and intricate reasoning tasks with impressive accuracy and throughput. By activating only a portion of its parameters for each token, this architecture significantly enhances computational efficiency while preserving robust reasoning capabilities, making it ideal for scalable inference under heavy workloads. The Nemotron-3 Super comprises approximately 120 billion parameters, with around 12 billion being active during inference, which substantially boosts its ability to handle multi-step reasoning and collaborative interactions among agents within extensive contexts. Such advancements make it a powerful tool for tackling diverse challenges in AI applications.

Carziqo is a cutting-edge technology firm dedicated to revolutionizing autonomous driving and smart mobility solutions. Our mission is to reshape how individuals travel and earn through advanced transportation innovations. As a worldwide frontrunner in the self-driving car rental market, Carziqo offers both individuals and businesses access to high-performance, intelligent, and secure autonomous vehicles, allowing everyone to seamlessly adopt the future of technology. We deliver more than merely a vehicle; we offer a comprehensive intelligent mobility ecosystem. With the Carziqo platform, customers can effortlessly rent autonomous cars for logistics services or ride-sharing, creating opportunities for additional income generation. This service caters to both independent entrepreneurs and corporate clients, enabling them to achieve a more efficient, environment-friendly, and economical approach to smart transport solutions. Ultimately, Carziqo is committed to enhancing the overall travel and earning experiences through innovation and advanced technology.

The Qualcomm® Snapdragon Ride™ Platform stands out as one of the most sophisticated, adaptable, and fully customizable automated driving systems in the automotive sector. It offers automotive manufacturers and suppliers the flexibility to implement the sought-after safety, convenience, and autonomous driving capabilities of today while maintaining the potential for future scalability. This platform boasts dependable, high-performance capabilities tailored for automotive needs, all while ensuring lower power consumption, enhanced simplicity, and greater safety in vehicles. Unlike many other autonomous driving technologies that depend on liquid cooling systems, the Snapdragon Ride Platform utilizes passive or air-cooling methods, making it a more efficient choice. With its unique multi-ECU aggregation feature, this versatile platform can seamlessly transition from active safety measures to convenience features and ultimately to complete self-driving solutions, accommodating a diverse array of vehicles. Furthermore, the Snapdragon Ride Autonomous Stack complements the high-performance, energy-efficient hardware, creating a powerful and sophisticated driving and perception system for vehicles today. This combination positions the platform as a leader in the realm of automotive innovation, paving the way for future advancements in the industry.

Momenta stands out as a premier company in the field of autonomous driving technology. Committed to transforming the landscape of mobility, Momenta delivers innovative solutions that facilitate various levels of driving autonomy. The company has established a distinctive and scalable roadmap towards achieving complete autonomous driving by integrating a data-centric methodology with the continuous refinement of algorithms, a strategy known as the “flywheel approach.” Additionally, Momenta employs a “two-leg” product strategy, which encompasses Mpilot, its highly autonomous driving solution ready for mass production, and MSD (Momenta Self-Driving), aimed at reaching full autonomy. Mpilot is specifically designed as a mass-production-ready software solution for automated driving in private vehicles. A key component of this offering is Mpilot X, which delivers a comprehensive and highly autonomous driving experience across all driving scenarios, featuring essential functionalities such as Mpilot Highway, Mpilot Urban, and Mpilot Parking. With a focus on innovation and user experience, Momenta is poised to lead the way in the future of transportation.

Kodiak AI focuses on its innovative Kodiak Driver, a comprehensive autonomous driving platform that merges sophisticated AI-driven software with adaptable, vehicle-independent hardware to facilitate scalable, practical autonomy for trucks and terrestrial vehicles. The system is crafted for seamless integration across various vehicle models and operating environments, utilizing a comprehensive array of sensors housed in interchangeable SensorPods for complete 360° awareness. It employs deep-learning perception algorithms to decipher complex surroundings, along with advanced planning features that predict road changes, while also incorporating backup systems for computing, power, steering, and braking designed for safety and dependability in high-demand scenarios. This technology is primed for implementation in commercial long-haul trucking, industrial logistics, and defense-related ground vehicles. Additionally, its connectivity and telematics capabilities support over-the-air updates, enable remote management of fleets, and include Assisted Autonomy features that permit human monitoring, enhancing the overall safety and efficiency of operations. Ultimately, Kodiak AI's solutions strive to redefine the future of transportation by ensuring both reliability and adaptability in autonomous systems.

Inkling-Small is an efficient Mixture-of-Experts transformer model built to provide performance comparable to Inkling while using a much smaller active parameter footprint. The model has 276 billion total parameters and 12 billion active parameters, making it designed for strong capability with more efficient compute usage. Inkling-Small was trained on NVIDIA GB300 NVL72 systems and supports native reasoning across text, images, and audio. It offers context windows of up to one million tokens, making it suitable for long documents, large codebases, multimodal context, and extended agent workflows. Users can set reasoning effort from minimal to extra high to control the balance between speed, cost, compute, and task complexity. The model benefits from improved pre-training data, post-training with on-policy distillation from Inkling, and extended agentic coding reinforcement learning. These improvements helped Inkling-Small surpass its larger counterpart on reasoning and coding benchmarks. Its encoder-free multimodal architecture processes audio as dMel spectrograms and images as 40-by-40-pixel patches alongside text tokens. By combining efficient MoE scaling, long-context reasoning, multimodal input, coding strength, and adjustable thinking effort, Inkling-Small is built for practical high-performance AI deployment.

NVIDIA's Isaac GR00T (Generalist Robot 00 Technology) serves as an innovative research platform aimed at the creation of versatile humanoid robot foundation models and their associated data pipelines. This platform features models such as Isaac GR00T-N, alongside synthetic motion blueprints, GR00T-Mimic for enhancing demonstrations, and GR00T-Dreams, which generates novel synthetic trajectories to expedite the progress in humanoid robotics. A recent highlight is the introduction of the open-source Isaac GR00T N1 foundation model, characterized by a dual-system cognitive structure that includes a rapid-response “System 1” action model and a language-capable, deliberative “System 2” reasoning model. The latest iteration, GR00T N1.5, brings forth significant upgrades, including enhanced vision-language grounding, improved following of language commands, increased adaptability with few-shot learning, and support for new robot embodiments. With the integration of tools like Isaac Sim, Lab, and Omniverse, GR00T enables developers to effectively train, simulate, post-train, and deploy adaptable humanoid agents utilizing a blend of real and synthetic data. This comprehensive approach not only accelerates robotics research but also opens up new avenues for innovation in humanoid robot applications.

As a leading provider of automated platform solutions globally, we offer a highly adaptable R&D vehicle platform that can significantly enhance your projects related to advanced driver assistance systems (ADAS), algorithm innovation, and autonomous driving initiatives, or elevate your driverless technology endeavors to new heights. You can methodically define the specifications of your R&D vehicle platform, which includes everything from the vehicle itself to its sensors, software, and data storage components. When you choose to purchase a platform from AutonomouStuff, you gain not just a product but a partnership; an experienced project manager will be assigned to you, ensuring consistent communication and keeping you informed about platform advancements, while also guaranteeing that your requirements are fully addressed. This collaborative approach allows us to adapt to your evolving needs throughout the development process.

NVIDIA's Nemotron 3 represents a collection of open large language models crafted to drive advanced reasoning, conversational AI, and autonomous AI agents. This series consists of three distinct models tailored for varying scales of AI workloads, all while ensuring remarkable efficiency and precision. Emphasizing "agentic AI" features, these models are capable of executing multi-step reasoning, collaborating with tools, and functioning as integral parts of multi-agent systems utilized across automation, research, and enterprise sectors. The underlying architecture employs a hybrid mixture-of-experts (MoE) approach paired with transformer techniques, enabling the activation of only specific parameter subsets for each task, thereby enhancing performance and minimizing computational expenses. Designed to excel in reasoning, dialogue, and strategic planning, the Nemotron 3 models are optimized for high throughput, making them suitable for extensive deployment across diverse applications. Additionally, their innovative architecture allows for greater adaptability and scalability, ensuring they meet the evolving demands of modern AI challenges.

Monitor your vehicles easily Make your driver's commute easier. Optimize your fleet usage. Our team is available to help you. We can help you organize your vehicle fleet. We offer a full range of mobility solutions, including onboard telematics and online supervision platforms. It's never been easier to manage a mobility service. Unlimited mobility for your drivers. Innovative all-inclusive service. For corporate or P2P use.

Selenium stands as our premier product, representing an extensive full-stack autonomy system developed through over 500 person-years of dedicated work. This comprehensive suite of software for vehicles, designed to operate with a drive-by-wire interface and minimal computing resources, enables complete autonomy for land-based vehicles. Selenium is capable of converting any compatible vehicle platform into an autonomous unit, whether for prototype development or mass production. Comprised of a series of interoperable software components, it equips the vehicle to effectively address three fundamental inquiries: Where am I? What surrounds me? What actions should I take next? Encompassing a wide range of technologies, Selenium includes everything from low-level device drivers to calibration, four-modal localization, mapping, perception, machine learning, and planning, with its impressive vertical integration extending to user interfaces and data export systems. Notably, it operates independently of GPS or HD-Maps, although these can still be integrated when available, thus enhancing its versatility and application in diverse environments. With this innovative technology, we are redefining the future of autonomous vehicles.

NODAR has revolutionized stereo vision through innovative algorithms that allow standard cameras to achieve extraordinary levels of 3D range, accuracy, and dependability. This cutting-edge 3D vision technology is meticulously designed to extend the boundaries of what can be accomplished in both automotive and industrial sectors. In contexts such as passenger vehicles, self-driving trains, and security monitoring, the availability of dependable 3D data is vital for ensuring safety and optimal performance. NODAR excels in providing top-tier 3D spatial information for autonomous systems spanning various industries. The shift toward autonomy is reshaping virtually every facet of modern life, enhancing productivity, offering convenience, and improving safety standards. Operating in outdoor environments often presents challenges, as automated machinery must function continuously in adverse conditions marked by harsh weather, strong glare, dust, and significant vibrations. NODAR's innovative technology and products furnish unmatched data quality, precision, and reliability, which are essential for safety-critical applications reliant on accurate 3D information. This commitment to excellence positions NODAR as a leader in providing solutions that meet the evolving needs of various sectors.

The NVIDIA Agent Toolkit is an extensive framework and solution stack that facilitates the creation, deployment, and scaling of autonomous AI agents capable of reasoning, planning, and executing intricate tasks within enterprise environments. In contrast to traditional generative AI that reacts to isolated prompts, agentic AI employs advanced reasoning and iterative planning methods to independently tackle multi-step challenges, empowering systems to analyze information, devise strategies, and carry out workflows without the need for constant human oversight. This toolkit encompasses various elements of the NVIDIA AI ecosystem, featuring pretrained models, microservices, and development frameworks, which enable organizations to develop context-aware AI agents that leverage their own data for optimal performance. These agents can effectively process substantial amounts of both structured and unstructured data sourced from enterprise systems, allowing them to understand context and synchronize actions across diverse applications for automating processes in areas such as customer support, software development, analytics, and operational workflows. Additionally, by enhancing collaboration among various business functions, the NVIDIA Agent Toolkit can significantly improve efficiency and decision-making across organizations.

Phi-4 is an advanced small language model (SLM) comprising 14 billion parameters, showcasing exceptional capabilities in intricate reasoning tasks, particularly in mathematics, alongside typical language processing functions. As the newest addition to the Phi family of small language models, Phi-4 illustrates the potential advancements we can achieve while exploring the limits of SLM technology. It is currently accessible on Azure AI Foundry under a Microsoft Research License Agreement (MSRLA) and is set to be released on Hugging Face in the near future. Due to significant improvements in processes such as the employment of high-quality synthetic datasets and the careful curation of organic data, Phi-4 surpasses both comparable and larger models in mathematical reasoning tasks. This model not only emphasizes the ongoing evolution of language models but also highlights the delicate balance between model size and output quality. As we continue to innovate, Phi-4 stands as a testament to our commitment to pushing the boundaries of what's achievable within the realm of small language models.

Laguna M.1 stands out as Poolside's most proficient model for agentic coding, meticulously developed in-house specifically for enhancing software development workflows. This model features a total of 225 billion parameters, utilizing a Mixture of Experts architecture with 23 billion activated parameters, and has been trained entirely within the organization on a dataset consisting of 30 trillion tokens, leveraging the power of 6,144 interconnected NVIDIA H200 GPUs. Poolside undertook the task of training Laguna M.1 from the ground up, employing its proprietary data, dedicated training codebase, and an asynchronous on-policy reinforcement learning approach within its agent framework, all tailored for agentic coding applications. The design of the model ensures optimal performance within Poolside's coding agent, enabling it to effectively reason through software tasks, interact with various tools, edit code, execute tests, and facilitate extended autonomous development sessions. Specifically crafted for developers and teams tackling intricate coding challenges, Laguna M.1 offers enhanced capabilities in reasoning, architectural comprehension, terminal operations, and multi-step execution, surpassing what lighter models can achieve. Ultimately, its robust feature set positions it as an essential asset for those engaged in demanding software projects.

We collaborate to develop innovative solutions that help our clients transition to sustainable mobility while propelling the automotive industry forward. Fueled by the integration of cutting-edge technologies such as AI, IoT, and connected infrastructure, alongside digitization and electrification, we are swiftly moving toward a transformed future in automotive. This pivotal moment promises unparalleled freedom characterized by zero accidents, zero emissions, and no ownership, presenting a remarkable opportunity for society, the economy, and the environment. Our role is to empower automakers to push the boundaries of crucial automotive and mobility technology advancements. By merging the agility and dynamic nature of a startup with the extensive reach and resources of a larger enterprise, we are able to provide complex solutions rapidly, even in critical situations. As we navigate the era of autonomous driving, we focus on addressing both current and future software demands. Moreover, we emphasize the importance of creating distinct and highly personalized in-vehicle experiences that leverage intelligent technology. This approach not only enhances the driving experience but also aligns with the broader goals of sustainability and innovation in the automotive sector.

We are advancing safe and dependable autonomous driving technology on a global scale. After logging millions of kilometers in complex scenarios during our autonomous road tests, we have established a robust groundwork for delivering scalable autonomous driving systems. In December 2018, Pony.ai pioneered the launch of its Robotaxi service, which allows passengers to summon self-driving vehicles using the PonyPilot+ App, initiating a new chapter in safe and enjoyable transportation. This service is currently operational in cities such as Guangzhou, Beijing, Irvine, CA, and Fremont, CA. Additionally, we have initiated autonomous mobility pilots in various locations throughout the United States and China, catering to hundreds of riders on a daily basis. These pilot programs have provided us with valuable insights and a solid technical and operational base to enhance and expand our offerings. United in our mission, we are addressing some of the most significant technological challenges in the mobility sector. Each day, we are making tangible advancements toward our vision of making autonomous mobility a universal reality. Our dedication to innovation drives us forward as we continue to strive for excellence in this evolving field.

A robust application platform designed for autonomous driving systems utilizes diverse collective data to perceive the outside environment and make informed driving decisions independently. This system is supported by a scalable distributed computing infrastructure that leverages many-core processors alongside a variety of processor types to deliver the enhanced computational power essential for sophisticated information processing. Furthermore, the platform is designed to be adaptable, enabling software applications to operate across a range of hardware resources while maintaining real-time capabilities to guarantee both safety and reliability in autonomous operations. This versatility ensures that the system can evolve alongside advancements in technology and changing operational requirements.

Mercury 2 represents a groundbreaking advancement in reasoning models, specifically designed for real-time voice interaction as it can quickly answer phone calls. Unlike traditional autoregressive models that leave callers in silence while generating responses one token at a time, Mercury 2 employs a diffusion large language model architecture capable of producing over 1000 tokens per second with standard NVIDIA GPUs. This remarkable speed allows it to complete a full reasoning process and begin speaking within a timeframe that aligns with natural conversational flow, effectively shortening the typical wait time from several seconds to approximately 300 milliseconds. The operational mechanism of Mercury models involves transforming clear text into noise, after which a conventional Transformer is trained to reverse this transformation and predict the original text across all positions at once. By utilizing a denoising approach that engages multiple tokens simultaneously, generation becomes more efficient, enabling speeds akin to custom silicon on NVIDIA H100s while improving responsiveness in voice applications. As a result, Mercury 2 not only enhances user experience but also sets a new standard for interactive voice technologies.

AfterQuery serves as a practical research platform aimed at generating high-quality training datasets for cutting-edge artificial intelligence models by emulating the cognitive processes of seasoned professionals as they think, reason, and tackle challenges in their fields. By converting real-world work scenarios into organized datasets, it provides insights that transcend mere outputs, incorporating intricate decision-making, trade-offs, and contextual reasoning that typical internet-sourced data fails to capture. The platform collaborates closely with subject matter experts to produce supervised fine-tuning data, which includes prompt–response pairs alongside comprehensive reasoning trails, in addition to reinforcement learning datasets featuring expertly crafted prompts and assessment frameworks that translate subjective evaluations into scalable reward mechanisms. Furthermore, it develops customized agent environments using various APIs and tools, facilitating the training and evaluation of models within realistic workflows while also tracking computer-use trajectories that illustrate how individuals engage with software in a detailed, step-by-step manner. This multi-faceted approach ensures that the data generated not only reflects expert insights but is also adaptable for a wide range of applications in the evolving landscape of artificial intelligence.

Applied Intuition's Vehicle OS is a highly adaptable and modular solution designed to assist automakers, commercial fleets, and defense contractors in creating, launching, and maintaining a wide array of vehicle software, hardware, and AI applications that span various areas, including advanced driver-assistance systems (ADAS), entertainment systems, autonomous driving, and digital services. The integrated on-board SDK offers a real-time operating system, essential drivers, middleware, and a foundational compute architecture tailored for both safety-critical and consumer-oriented applications, while the external platform facilitates cloud-based data logging, remote diagnostics, over-the-air (OTA) updates, and management of digital twins. Developers benefit from a comprehensive Workbench environment that includes built-in build and testing resources, continuous integration pipelines, and automated validation processes. This platform effectively connects vehicle intelligence across different ecosystems by merging autonomy frameworks, simulation tools for vehicle dynamics and sensor interactions, and a robust ecosystem of developer tools. By providing these capabilities, it empowers developers to innovate and enhance the future of transportation.

The Linker VisionAI Platform offers a holistic, all-in-one solution for vision AI, incorporating elements of simulation, training, and deployment to enhance the capabilities of smart cities and businesses. It is built around three essential components: Mirra, which generates synthetic data through NVIDIA Omniverse and NVIDIA Cosmos; DataVerse, which streamlines data curation, annotation, and model training with NVIDIA NeMo and NVIDIA TAO; and Observ, designed for the deployment of large-scale Vision Language Models (VLM) using NVIDIA NIM. This cohesive strategy facilitates a smooth progression from simulated data to practical application, ensuring that AI models are both resilient and flexible. By utilizing urban camera networks and advanced AI technologies, the Linker VisionAI Platform supports a variety of functions, such as managing traffic, enhancing worker safety, and responding to disasters. In addition, its comprehensive capabilities allow organizations to make well-informed decisions in real-time, significantly improving operational efficiency across diverse sectors.

Carver21 serves as a foundational framework for smart vehicles, designed to effectively adapt to your specific perception requirements, whether it’s enhancing safety mechanisms or facilitating self-driving capabilities. This innovative system ensures that advanced automotive technologies can evolve alongside user needs.

The MAI-1 Preview marks the debut of Microsoft AI's fully in-house developed foundation model, utilizing a mixture-of-experts architecture for streamlined performance. This model has undergone extensive training on around 15,000 NVIDIA H100 GPUs, equipping it to adeptly follow user instructions and produce relevant text responses for common inquiries, thus illustrating a prototype for future Copilot functionalities. Currently accessible for public testing on LMArena, MAI-1 Preview provides an initial look at the platform's direction, with plans to introduce select text-driven applications in Copilot over the next few weeks aimed at collecting user insights and enhancing its capabilities. Microsoft emphasizes its commitment to integrating its proprietary models, collaborations with partners, and advancements from the open-source sector to dynamically enhance user experiences through millions of distinct interactions every day. This innovative approach illustrates Microsoft's dedication to continuously evolving its AI offerings.

GLM-5 is a next-generation open-source foundation model from Z.ai designed to push the boundaries of agentic engineering and complex task execution. Compared to earlier versions, it significantly expands parameter count and training data, while introducing DeepSeek Sparse Attention to optimize inference efficiency. The model leverages a novel asynchronous reinforcement learning framework called slime, which enhances training throughput and enables more effective post-training alignment. GLM-5 delivers leading performance among open-source models in reasoning, coding, and general agent benchmarks, with strong results on SWE-bench, BrowseComp, and Vending Bench 2. Its ability to manage long-horizon simulations highlights advanced planning, resource allocation, and operational decision-making skills. Beyond benchmark performance, GLM-5 supports real-world productivity by generating fully formatted documents such as .docx, .pdf, and .xlsx files. It integrates with coding agents like Claude Code and OpenClaw, enabling cross-application automation and collaborative agent workflows. Developers can access GLM-5 via Z.ai’s API, deploy it locally with frameworks like vLLM or SGLang, or use it through an interactive GUI environment. The model is released under the MIT License, encouraging broad experimentation and adoption. Overall, GLM-5 represents a major step toward practical, work-oriented AI systems that move beyond chat into full task execution.

Tronis serves as a platform for virtual prototyping and for ensuring the safety of driver assistance technologies, particularly in the realm of highly automated or fully autonomous vehicles. Utilizing an advanced 3D game engine, it allows for the effective simulation of real-world driving conditions and traffic situations, which can be utilized for testing purposes, such as assessing camera and radar-based environmental perception. This innovative approach can greatly expedite the development process of these systems while minimizing the reliance on physical prototypes. Consequently, it offers a cost-effective solution that enhances the efficiency of creating safer autonomous driving technologies.

Olmo 3 represents a comprehensive family of open models featuring variations with 7 billion and 32 billion parameters, offering exceptional capabilities in base performance, reasoning, instruction, and reinforcement learning, while also providing transparency throughout the model development process, which includes access to raw training datasets, intermediate checkpoints, training scripts, extended context support (with a window of 65,536 tokens), and provenance tools. The foundation of these models is built upon the Dolma 3 dataset, which comprises approximately 9 trillion tokens and utilizes a careful blend of web content, scientific papers, programming code, and lengthy documents; this thorough pre-training, mid-training, and long-context approach culminates in base models that undergo post-training enhancements through supervised fine-tuning, preference optimization, and reinforcement learning with accountable rewards, resulting in the creation of the Think and Instruct variants. Notably, the 32 billion Think model has been recognized as the most powerful fully open reasoning model to date, demonstrating performance that closely rivals that of proprietary counterparts in areas such as mathematics, programming, and intricate reasoning tasks, thereby marking a significant advancement in open model development. This innovation underscores the potential for open-source models to compete with traditional, closed systems in various complex applications.

micro1 Intelligence is a data research company focused on advancing frontier artificial intelligence through expert human data, contextual evaluations, and realistic training environments. The company develops infrastructure that enables AI models and autonomous agents to learn from high-quality human expertise rather than relying solely on synthetic or conventional datasets. Its Realm platform creates reinforcement learning environments that simulate real-world scenarios to generate human feedback and improve agent reasoning capabilities. Cortex provides contextual evaluation tools that measure and optimize AI agent performance in production environments using realistic tasks and benchmarks. micro1 also develops robotics datasets that capture high-fidelity real-world interactions to support the training of next-generation embodied AI systems. Alongside its technology platforms, the company publishes research on AI benchmarking, extraction systems, pathology reasoning, human data markets, and model evaluation. Expert opportunities and data partnerships allow professionals and organizations to contribute specialized knowledge that improves AI training quality. By combining research, human expertise, and production-focused evaluation, micro1 Intelligence helps accelerate the development of safer, more capable AI systems. The platform is designed to support frontier AI companies building intelligent agents, reasoning systems, and robotics applications with higher-quality training data.

Engineered using top-tier hardware and innovative software, the Aurora Driver is crafted to be versatile across different vehicle categories and applications, thereby enabling the advantages of autonomous driving in various sectors, such as freight transportation, local delivery services, and passenger transport. This advanced system incorporates sensors that gather environmental data, software that devises a secure route, and a computer that integrates these components with the vehicle itself. Capable of handling any vehicle, from compact cars to heavy-duty trucks, the Aurora Driver is equipped with the Aurora Computer, which acts as the core interface connecting our hardware with the autonomy software, facilitating smooth integration with diverse vehicle types. Our tailor-made sensor suite, which includes FirstLight Lidar, advanced long-range radars, and high-definition cameras, collaborates effectively to create a three-dimensional view of the surroundings, providing the Aurora Driver with a comprehensive 360˚ awareness of its environment in real time. This sophisticated technology ensures that the Aurora Driver can adapt to varying driving conditions, enhancing safety and efficiency across all its applications.

We are developing Hivemind, an AI pilot that empowers swarms of drones and aircraft to function independently, without the need for GPS, communication systems, or direct human oversight. Our aim is to enhance the safety of both military personnel and civilians through advanced intelligent systems. Unlike simple pre-programmed behaviors and fixed waypoints, Hivemind mimics human pilots by interpreting and responding to the complexities of the battlefield, making real-time decisions without relying on GPS or prior instructions. This groundbreaking technology is the first fully autonomous AI pilot to be deployed in combat operations since 2018. Hivemind's capabilities extend from conducting indoor reconnaissance with quadcopters to facilitating coordinated air defense missions with fixed-wing drones and engaging in F-16 dogfights. As it learns and autonomously carries out a variety of missions, Hivemind represents a pioneering evolution in aerial warfare, ensuring sustained aerial superiority across land, air, and sea, especially in high-stakes tactical environments. This innovative approach marks a significant advancement in military technology and operational efficiency.