NVIDIA has introduced the open Llama Nemotron family of models, equipped with advanced reasoning capabilities to offer developers and enterprises a robust foundation for building AI agents that operate autonomously or collaboratively to tackle complex challenges.

Built on Llama models, the NVIDIA Llama Nemotron reasoning family delivers on-demand AI reasoning capabilities. NVIDIA enhanced the new reasoning model family during post-training to improve multistep math, coding, reasoning and complex decision-making.

This refinement process boosts accuracy of the models by up to 20% compared with the base model and optimizes inference speed by 5x compared with other leading open reasoning models. The improvements in inference performance mean the models can handle more complex reasoning tasks, enhance decision-making capabilities and reduce operational costs for enterprises.

Leading agent AI platform pioneers — including Accenture, Amdocs, Atlassian, Box, Cadence, CrowdStrike, Deloitte, IQVIA, Microsoft, SAP and ServiceNow — are collaborating with NVIDIA on its new reasoning models and software.

“Reasoning and agentic AI adoption is incredible,”

said Jensen Huang, founder and CEO of NVIDIA.

“NVIDIA’s open reasoning models, software and tools give developers and enterprises everywhere the building blocks to create an accelerated agentic AI workforce.”

NVIDIA Post-Training Boosts Accuracy and Reliability for Enterprise Reasoning

Built to deliver production-ready AI reasoning, the Llama Nemotron model family is available as NVIDIA NIM™ microservices in Nano, Super and Ultra sizes — each optimized for different deployment needs.

The Nano model delivers the highest accuracy on PCs and edge devices, the Super model offers the best accuracy and highest throughput on a single GPU, and the Ultra model will provide maximum agentic accuracy on multi-GPU servers.

NVIDIA conducted extensive post-training on NVIDIA DGX™ Cloud using high-quality curated synthetic data generated by NVIDIA Nemotron™ and other open models, as well as additional curated datasets cocreated by NVIDIA.

The tools, datasets and post-training optimization techniques used to develop the models will be openly available, giving enterprises the flexibility to build their own custom reasoning models.

Agentic Platforms Team With NVIDIA to Enhance Reasoning for Industries

Agentic AI platform industry leaders are working with the Llama Nemotron reasoning models to deliver advanced reasoning to enterprises.

Microsoft is integrating Llama Nemotron reasoning models and NIM microservices into Microsoft Azure AI Foundry. This expands the Azure AI Foundry model catalog with options for customers to enhance services like Azure AI Agent Service for Microsoft 365.

SAP is tapping Llama Nemotron models to advance SAP Business AI solutions and Joule, the AI copilot from SAP. Additionally, it is using NVIDIA NIM and NVIDIA NeMo™ microservices to promote increased code completion accuracy for SAP ABAP programming language models.

“We are collaborating with NVIDIA to integrate Llama Nemotron reasoning models into Joule to enhance our AI agents, making them more intuitive, accurate and cost effective,”

said Walter Sun, global head of AI at SAP.

“These advanced reasoning models will refine and rewrite user queries, enabling our AI to better understand inquiries and deliver smarter, more efficient AI-powered experiences that drive business innovation.”

ServiceNow is harnessing Llama Nemotron models to build AI agents that offer greater performance and accuracy to enhance enterprise productivity across industries.

Accenture has made NVIDIA Llama Nemotron reasoning models available on its AI Refinery platform — including new industry agent solutions announced today — to enable clients to rapidly develop and deploy custom AI agents tailored to industry-specific challenges, accelerating business transformation.

Deloitte is planning to incorporate Llama Nemotron reasoning models into its recently announced Zora AI agentic AI platform designed to support and emulate human decision-making and action with agents that include deep functional- and industry-specific business knowledge and built-in transparency.

NVIDIA AI Enterprise Delivers Essential Tools for Agentic AI

Developers can deploy NVIDIA Llama Nemotron reasoning models with new NVIDIA agentic AI tools and software to streamline the adoption of advanced reasoning in collaborative AI systems.

All part of the NVIDIA AI Enterprise software platform, the latest agentic AI building blocks include:

• The NVIDIA AI-Q Blueprint, which enables enterprises to connect knowledge to AI agents that can autonomously perceive, reason and act. Built with NVIDIA NIM microservices, the blueprint integrates NVIDIA NeMo Retriever™ for multimodal information retrieval and enables agent and data connections, optimization and transparency using the open-source NVIDIA AgentIQ toolkit.
• The NVIDIA AI Data Platform, a customizable reference design for a new class of enterprise infrastructure with AI query agents built with the AI-Q Blueprint.
• New NVIDIA NIM microservices, which optimize inference for complex agentic AI applications and enable continuous learning and real-time adaptation across any environment. The microservices ensure reliable deployment of the latest models from leading model builders including Meta, Microsoft and Mistral AI.
• NVIDIA NeMo microservices, which provide an efficient, enterprise-grade solution to quickly establish and maintain a robust data flywheel that enables AI agents to continuously learn from human- and AI-generated feedback. The NVIDIA AI Blueprint for building a data flywheel will offer a reference architecture for developers to easily build and optimize data flywheels using NVIDIA microservices.

The NVIDIA Llama Nemotron Nano and Super models and NIM microservices are available as a hosted application programming interface from build.nvidia.com and Hugging Face. Access for development, testing and research is free for members of the NVIDIA Developer Program.

Enterprises can run Llama Nemotron NIM microservices in production with NVIDIA AI Enterprise on accelerated data center and cloud infrastructure. Developers can sign up to be notified when NVIDIA NeMo microservices are publicly available.

The NVIDIA AI-Q Blueprint is expected to be available in April. The NVIDIA AgentIQ toolkit is available now on GitHub.

All these facts are well known to experienced people in the tech industry and were cited by Mark Zuckerberg in his article ‘Open Source AI Is the Path Forward’ posted on Meta’s blog.

Open-source AI is rapidly closing the gap

According to Zuckerburg, AI is expected to develop in a similar manner. Currently, several tech companies are developing leading closed models. However, open-source AI is rapidly closing the gap. Last year, Llama 2 was comparable to an older generation of models. This year, Llama 3 is competitive with the most advanced models and leading in some areas. Starting next year, future Llama models are expected to become the most advanced in the industry. Even now, Llama leads in openness, modifiability, and cost efficiency.

Llama 3.1 405B, the first frontier-level open-source AI model

Significant steps are being taken toward making open-source AI the industry standard. Llama 3.1 405B, the first frontier-level open-source AI model, is being released along with new and improved Llama 3.1 70B and 8B models. These models offer significantly better cost/performance compared to closed models. The 405B model’s openness makes it ideal for fine-tuning and distilling smaller models.

In addition to releasing these models, there is collaboration with companies to grow the broader ecosystem. Amazon, Databricks, and NVIDIA are launching full suites of services to support developers in fine-tuning and distilling their models. Innovators like Groq have built low-latency, low-cost inference serving for the new models. The models will be available on major clouds, including AWS, Azure, Google, Oracle, and more. Companies like Scale.AI, Dell, and Deloitte are ready to help enterprises adopt Llama and train custom models with their data. As the community grows and more companies develop new services, Llama can collectively become the industry standard, bringing AI benefits to everyone.

“Meta is committed to open-source AI”

Mark Zuckerberg is claiming that Meta is committed to open-source AI. He even outlines why open source is the best development stack, why open sourcing Llama is beneficial for Meta, and why open-source AI is good for the world, ensuring long-term viability.

Why Open Source AI Is Good for Developers

Open source AI is beneficial for developers as it allows for customization of models to meet specific organizational needs without external oversight, provides independence from closed vendors ensuring control over models and flexibility, enhances data security by allowing models to run locally with transparent development processes, offers cost efficiency with models like Llama 3.1 405B running at approximately half the cost of closed models, and represents a promising future investment as open source technology is advancing rapidly and is expected to remain advantageous in the long term.

You can access the models now at llama.meta.com. 

Photo: about.meta.com

A group of technology companies, including Intel, Google, Arm, Qualcomm and Samsung, created the Unified Accelerator Fund (UXL) alliance, through which they will work together to remove the advantages that NVIDIA has as a dominant player in the artificial intelligence market, reports The Verge.

The alliance will undertake the mission of developing an open-source software package that will allow AI professionals to break away from NVIDIA’s technology. Additionally, the code that developers create will be able to function on any machine with any chip.

The Unified Acceleration Foundation is still in the design phase, but is expected to begin operations by the beginning of the second half of the year. UXL currently supports the OneAPI open standard. It was created by the Intel company.

Members of the alliance believe that it will remove restrictions on developers using specific programming languages, code sets and any other tools intended to bind developers to the use of certain architectures. For example – NVIDIA’s CUDA platform, which strives for this.

After focusing entirely on making chips for AI models like the H100, the company reached $2 trillion in market capitalization last month. This secured NVIDIA’s top spot among chip makers. The company is also planning to release new AI models soon – H200 GPU.

The chips that NVIDIA manufactures require developers to use the CUDA architecture, which supersedes all other chip manufacturers’ architectures by leaps and bounds. Jensen Huang, CEO of NVIDIA, also shared that 4 million developers use CUDA.

However, technology companies are not giving up and continue to try to create alternative architectures of their own to displace the dominant one in the market.

For now, the creators of UXL intend to support NVIDIA hardware and code. Initially, the project will aim to develop options for AI applications.

The group of tech companies that make up the alliance is looking to collaborate with other powerhouses like Microsoft and Amazon to make sure the solutions they provide can be built into any chip or hardware.

Vultr Cloud Inference offers customers scalability, lower latency and delivers cost efficiencies, the company said in the release.

Kevin Cochrane, chief marketing officer at Vultr says of the new platform that Vultr Cloud Inference provides a technology foundation with which organizations can deploy AI models globally, providing low-latency access and a consistent user experience across the world.

Vultr’s global infrastructure is powered by NVIDIA GPUs. With dedicated computer clusters available on six continents, Vultr Cloud Inference ensures that companies can comply with local data sovereignty, data residency, and privacy regulations by deploying their AI applications in regions that align with legal requirements and business objectives.

“The introduction of Vultr Cloud Inference will empower businesses to seamlessly integrate and deploy AI models trained on NVIDIA GPU infrastructure, helping them scale their AI applications globally”, said Matt McGrigg, director of global business development, cloud partners at NVIDIA.

With Vultr Cloud Inference, users can integrate and deploy their own models – regardless of the platforms on which they have been trained – into the Vultr infrastructure powered by NVIDIA GPUs.

“To harness the incredible potential of generative AI, enterprises need adaptable, scalable hybrid infrastructure in their local markets to bring AI supercomputing to their data. Our new service provides customers a fast and cost-effective way to adopt advanced AI infrastructure that’s operated and managed by experts globally”, said Charles Meyers, president and CEO of Equinix.

The service is available now and includes NVIDIA DGX systems, NVIDIA networking and the NVIDIA AI Enterprise software platform. Equinix installs and operates each customer’s privately owned NVIDIA infrastructure and can deploy services on their behalf in key International Business Exchange™ (IBX) data centers globally.

“Generative AI is transforming every industry. Now, enterprises can own NVIDIA AI supercomputing and software, paired with the operational efficiency of Equinix management, in hundreds of data centers worldwide”, said Jensen Huang, founder and CEO of NVIDIA.

Scalable Service to Fuel Innovation Across Industries

Through the service, enterprises can scale their infrastructure operations to achieve the level of AI performance needed to develop and run massive models.

Early access companies using the service include leaders in biopharma, financial services, software, automotive and retail, which are building AI Centers of Excellence to provide a strategic foundation for a broad range of rapidly developing LLM use cases.

These include accelerating time to market for new medications, developing AI copilots for customer service agents and building virtual productivity assistants.

Easy Access to Privately Managed NVIDIA AI Supercomputing

Equinix’s fully managed NVIDIA AI supercomputing service enables customers to operate their AI infrastructure in close proximity to their data.

The service offers high-speed private network access to global network service providers, enabling quick generative AI information retrieval across corporate wide area networks. In addition, it provides private, high-bandwidth interconnections to cloud services and enterprise service providers to facilitate AI workloads while meeting data security and compliance requirements.

Using the service, customers can easily access their NVIDIA AI Enterprise software to streamline the development and deployment of production-grade AI applications, including generative AI. NVIDIA AI Enterprise includes pretrained models, optimized frameworks and accelerated data science software libraries—such as the NVIDIA NeMo™ framework for building LLMs, NVIDIA RAPIDS™ for data science, NVIDIA Clara for healthcare and NVIDIA TensorRT™–LLM for performance optimization of large language models.

Enterprise-Level Support

The comprehensive solution also features enterprise-grade support and security. This includes Equinix’s IBX data center professionals to help customers rapidly build and deploy their custom AI models, as well as access to NVIDIA AI experts.

The fully managed private cloud service is available today.

Read more:
1. Less than Two Months to the FISSION International Science Festival
2. Android Releases First Preview of Android 15 for Developers
3. Kong Introduces Six New AI Plugins in Kong Gateway 3.6

To learn how to utilize these new AI features in Power Apps, users can refer to the following resources:

• Build apps through conversation (preview)
• Add Chatbot control to a canvas app (preview)
• Add Copilot control to a canvas app (preview)
• Leverage Azure OpenAI Service in AI Builder (preview)

It is important to note that these capabilities are currently in gated preview, and interested parties must apply for consideration to participate in the trial via the Limited preview request.

Preview features are not intended for production use and may have limited functionality. Customers are granted early access to these features to provide feedback before the official release. More information can be found in the preview terms.

The AI features in MS Power Apps are powered by Azure OpenAI Service and are in the process of being rolled out. The availability of these features may vary by region and could be subject to usage limits or capacity throttling.

To access the waitlist for this preview, users must meet certain prerequisites, including having their environment set in the United States region, an en-us browser language, a Microsoft Dataverse database, and a prioritized license. AI Builder must be enabled for the user’s environment to use AI models or controls leveraging AI models.

GitHub Copilot is powered by OpenAI GPT-3, a language prediction technology that generates a script that resembles human language. Today, we have chosen to present you with a variety of options that can replace GitHub Copilot, according to India AI.

Interesting and Good Alternatives to GitHub Copilot

Tabnine
Tabnine (formerly known as Codota) was among the first code completion tools that were introduced to the market. The Tabnine plugin for your preferred IDE supports the most popular programming languages, libraries and frameworks. Tabnine’s AI models are trained exclusively on open-source enabled code, ensuring that your work remains yours.

Tabnine supports widely used programming languages such as Typescript, Python, Rust, and Go. Additionally, each pattern in Tabnine is set up for specific languages, allowing for accurate autocompletion. When it comes to privacy and compliance, Tabnine always protects your code. Additionally, it integrates IDE tools such as VSCode, IntelliJ, Pycharm, Sublime, Rider, WebStorm, and AppCode.

CaptainStack
Captain Stack is a free and open-source Visual Studio Code plugin that combines the two. It is a code recommendation tool inspired by Copilot that uses Google instead of AI. It submits your search query to Google and retrieves and auto-completes replies from StackOverflow and Github Gist.

GPT-Code-Clippy (GPT-CC)
GPT-Code-Clippy is a collaborative effort to create GPT-Codex, an open-source counterpart of GitHub Copilot, a GPT-3-based AI pair programmer. It makes it easy for academics to evaluate large deep-learning models trained on code to determine their strengths and limitations. The GPT-Neo model, pre-trained on the Pile dataset, is the basic language model for GPT-CC. The model is trained with the Causal Language Modeling objective in mind.

IntelliCode
IntelliCode is an experimental AI coding helper trained on a selection of GitHub projects; it is a Visual Studio-exclusive Microsoft offering. Team completion is one of the most attractive features of IntelliCode. Tabnine supports this capability for all popular IDEs if you’re searching for an IDE-agnostic solution that enables team autocompletion training.

Asm-Dude
Asm-Dude is an addon for Microsoft Visual Studio that provides syntax highlighting and code completion for assembly files and disassembly windows. The primary features are syntax highlighting and descriptions, documentation links, code completion, code folding, structure help, and label analysis.

Kite
Kite is an AI-powered programming helper that assists programmers in writing Python code within JupyterLab. The platform enables developers to write faster by conserving keystrokes and displaying the appropriate data at the right moment.

Stability AI will work with AWS to make its open source tools and model available to more students, researchers, startups and enterprises.

“At Stability AI, our mission is to build the foundation to activate humanity’s potential through AI. AWS has played an integral role in scaling our open-source foundation models across modalities. We are delighted to run these models on Amazon SageMaker to enable thousands of developers and millions of users to leverage the power of AI with a robust set of tools”

 

said Emad Mostaque, founder and CEO of Stability AI.

 

“We look forward to seeing the amazing things that developers build and customers design and implement using collective intelligence and augmented technology”

he continued.

Stability AI plan to use AWS’ SageMaker ML platform, on top of its lower-level infrastructure services with GPUs and AWS’ own Trainium chips. Stability AI’s open source approach seems to be winning in bringing on more developers and driving innovation in this space.

Singh will lead SambaNova’s existing software development team, bringing two decades of experience heading up AI initiatives at startups, unicorns and top technology companies. Prior to joining SambaNova, Singh served as Head of AI Products at UiPath, where he founded the company’s AI product team and led the development of the AI-powered process automation platform. He previously spent 10 years at Microsoft and served as Head of Product for its Sales Intelligence AI solution. Singh said:

“I am beyond excited for the opportunity to join SambaNova and lead a dream team of engineers building a fully integrated stack that accelerates deployments and enables customers to undergo AI transformation overnight. As we sit on the cusp of a profound technological revolution, SambaNova is driving the industry and its customers forward into the next era of AI. I’m honoured to be a part of that vision.”

With years of experience scaling startups and small teams, Singh will further strengthen SambaNova’s world-class leadership team. In recent months, SambaNova has continued to add to its depth of industry experts, including the addition of CMO Amy Love and advisory board member Wade Shen, who both joined the company in July. Rodrigo Liang, co-founder and CEO of SambaNova, commented:

“SambaNova continues to attract and retain the industry’s top talent, and Prabhdeep joins our rapidly growing team as a highly valued addition. As we invest in a full-stack solution that meets the complex needs of our customers, Prabhdeep brings unmatched expertise in software delivering AI solutions in a wide range of verticals and scaling up startups into industry powerhouses.”

With AI becoming a business necessity in the global economy, customers need complete AI solutions that can run at scale in a financially viable way. With an integrated full-stack system, including best-in-class AI models, software and hardware, SambaNova provides the most expansive, accessible and impactful AI applications in the world — unleashing AI for everyone, everywhere.

SambaNova’s flagship Data Flow-as-a-Service, an extensible AI services platform, enables organizations to jump-start AI initiatives overnight by augmenting existing capabilities and staffing with a simple subscription. The platform is powered by DataScale, an integrated software and hardware platform delivering unrivalled performance, accuracy, scale and ease of use built on SambaNova’s Systems Reconfigurable Dataflow Architecture.

SambaNova continues to garner accolades throughout the industry, including recognition by Gartner as a Cool Vendor in its “Cool Vendors in AI Semiconductors” report, and industry awards for Best AI Product in Next-Generation Infrastructure by CogXand VentureBeat’s Innovation in Edge Award for 2021. The company was named one of CRN’s 10 Hottest AI Chip Makers in 2021 and one of CRN’s 20 Coolest Tech Startups Of 2020. SambaNova was also a Great Place to Work-Certified in 2021.

The work, shown at the automaker’s live-streamed AI Day, demonstrates the company’s continuing pursuit of vertical integration.

The D1 chip, part of Tesla’s Dojo supercomputer system, uses a 7-nanometer manufacturing process, with 362 teraflops of processing power, said Ganesh Venkataramanan, senior director of Autopilot hardware. Tesla places 25 of these chips onto a single “training tile,” and 120 of these tiles come together across several server cabinets, amounting to over an exaflop of power, Venkataramanan said:

“We are assembling our first cabinets pretty soon,” said Venkataramanan, who previously worked at chipmaker AMD.

He said the Tesla technology will be the fastest AI-training computer. Chipmaker Intel, graphics card maker Nvidia and start-up Graphcore are among the companies that make chips that companies can use to train AI models. The chips can help train models for recognizing a variety of items from video feeds collected by cameras inside Tesla vehicles. Model training requires extensive computing work. CEO Elon Musk said

“We should have Dojo operational next year.”

The work comes two years after Tesla began producing vehicles containing AI chips it built in house. Those chips help the car’s onboard software make decisions very quickly in response to what’s happening on the road.

Tesla currently offers what it calls a “Full Self-Driving Capability” add-on for new vehicles. The $10,000 package lets the car automatically change lanes, navigate on highways, move into parking spots and emerge from a parking spot to arrive by the driver. The Tesla website says later this year the package will also include the ability for a Tesla to automatically steer on city streets, although Tesla had previously promised that feature would come out in 2019.

Tesla’s website says Full-Self Driving elements “require active driver supervision and do not make the vehicle autonomous.” Earlier this year Tesla introduced a $199 monthly subscription for Tesla owners who wish to access Full-Self Driving.

Critics have said that Tesla’s moniker for its driver-assistance features is misleading. Tesla’s software does not offer Level 5 autonomy, where a car can completely drive in all circumstances without human intervention. Last year, a German court ruled that Tesla had misled consumers about the autonomous capabilities of its cars. The National Highway Traffic Safety Administration announced an investigation of Tesla’s Autopilot automatic steering, accelerating and braking capability earlier this week, following a string of crashes.