Nebius Token Factory Integrations

DeepSeek stands out as a state-of-the-art AI assistant, leveraging the sophisticated DeepSeek-V3 model that boasts an impressive 600 billion parameters for superior performance. Created to rival leading AI systems globally, it delivers rapid responses alongside an extensive array of features aimed at enhancing daily tasks' efficiency and simplicity. Accessible on various platforms, including iOS, Android, and web, DeepSeek guarantees that users can connect from virtually anywhere. The application offers support for numerous languages and is consistently updated to enhance its capabilities, introduce new language options, and fix any issues. Praised for its smooth functionality and adaptability, DeepSeek has received enthusiastic reviews from a diverse user base around the globe. Furthermore, its commitment to user satisfaction and continuous improvement ensures that it remains at the forefront of AI technology.

Mistral AI stands out as an innovative startup in the realm of artificial intelligence, focusing on open-source generative solutions. The company provides a diverse array of customizable, enterprise-level AI offerings that can be implemented on various platforms, such as on-premises, cloud, edge, and devices. Among its key products are "Le Chat," a multilingual AI assistant aimed at boosting productivity in both personal and professional settings, and "La Plateforme," a platform for developers that facilitates the creation and deployment of AI-driven applications. With a strong commitment to transparency and cutting-edge innovation, Mistral AI has established itself as a prominent independent AI laboratory, actively contributing to the advancement of open-source AI and influencing policy discussions. Their dedication to fostering an open AI ecosystem underscores their role as a thought leader in the industry.

DeepSeek-V3 represents a groundbreaking advancement in artificial intelligence, specifically engineered to excel in natural language comprehension, sophisticated reasoning, and decision-making processes. By utilizing highly advanced neural network designs, this model incorporates vast amounts of data alongside refined algorithms to address intricate problems across a wide array of fields, including research, development, business analytics, and automation. Prioritizing both scalability and operational efficiency, DeepSeek-V3 equips developers and organizations with innovative resources that can significantly expedite progress and lead to transformative results. Furthermore, its versatility makes it suitable for various applications, enhancing its value across industries.

Qwen is a next-generation AI system that brings advanced intelligence to users and developers alike, offering free access to a versatile suite of tools. Its capabilities include Qwen VLo for image generation, Deep Research for multi-step online investigation, and Web Dev for generating full websites from natural language prompts. The “Thinking” engine enhances Qwen’s reasoning and logical clarity, helping it tackle complex technical, analytical, and academic challenges. Qwen’s intelligent Search mode retrieves web information with precision, using contextual understanding and smart filtering. Its multimodal processing allows it to interpret content across text, images, audio, and video, enabling more accurate and comprehensive responses. Qwen Chat makes these features accessible to everyone, while developers can tap into the Qwen API to build apps, integrate Qwen into workflows, or create entirely new AI-driven experiences. The API follows an OpenAI-compatible format, making migration and adoption seamless. With broad platform support—web, Windows, macOS, iOS, and Android—Qwen delivers a unified, powerful AI ecosystem for all kinds of users.

DeepSeek-R1 is a cutting-edge open-source reasoning model created by DeepSeek, aimed at competing with OpenAI's Model o1. It is readily available through web, app, and API interfaces, showcasing its proficiency in challenging tasks such as mathematics and coding, and achieving impressive results on assessments like the American Invitational Mathematics Examination (AIME) and MATH. Utilizing a mixture of experts (MoE) architecture, this model boasts a remarkable total of 671 billion parameters, with 37 billion parameters activated for each token, which allows for both efficient and precise reasoning abilities. As a part of DeepSeek's dedication to the progression of artificial general intelligence (AGI), the model underscores the importance of open-source innovation in this field. Furthermore, its advanced capabilities may significantly impact how we approach complex problem-solving in various domains.

Mistral 7B is a language model with 7.3 billion parameters that demonstrates superior performance compared to larger models such as Llama 2 13B on a variety of benchmarks. It utilizes innovative techniques like Grouped-Query Attention (GQA) for improved inference speed and Sliding Window Attention (SWA) to manage lengthy sequences efficiently. Released under the Apache 2.0 license, Mistral 7B is readily available for deployment on different platforms, including both local setups and prominent cloud services. Furthermore, a specialized variant known as Mistral 7B Instruct has shown remarkable capabilities in following instructions, outperforming competitors like Llama 2 13B Chat in specific tasks. This versatility makes Mistral 7B an attractive option for developers and researchers alike.

Introducing Mistral NeMo, our latest and most advanced small model yet, featuring a cutting-edge 12 billion parameters and an expansive context length of 128,000 tokens, all released under the Apache 2.0 license. Developed in partnership with NVIDIA, Mistral NeMo excels in reasoning, world knowledge, and coding proficiency within its category. Its architecture adheres to industry standards, making it user-friendly and a seamless alternative for systems currently utilizing Mistral 7B. To facilitate widespread adoption among researchers and businesses, we have made available both pre-trained base and instruction-tuned checkpoints under the same Apache license. Notably, Mistral NeMo incorporates quantization awareness, allowing for FP8 inference without compromising performance. The model is also tailored for diverse global applications, adept in function calling and boasting a substantial context window. When compared to Mistral 7B, Mistral NeMo significantly outperforms in understanding and executing detailed instructions, showcasing enhanced reasoning skills and the ability to manage complex multi-turn conversations. Moreover, its design positions it as a strong contender for multi-lingual tasks, ensuring versatility across various use cases.

Qwen2.5 represents a state-of-the-art multimodal AI system that aims to deliver highly precise and context-sensitive outputs for a diverse array of uses. This model enhances the functionalities of earlier versions by merging advanced natural language comprehension with improved reasoning abilities, creativity, and the capacity to process multiple types of media. Qwen2.5 can effortlessly analyze and produce text, interpret visual content, and engage with intricate datasets, allowing it to provide accurate solutions promptly. Its design prioritizes adaptability, excelling in areas such as personalized support, comprehensive data analysis, innovative content creation, and scholarly research, thereby serving as an invaluable resource for both professionals and casual users. Furthermore, the model is crafted with a focus on user engagement, emphasizing principles of transparency, efficiency, and adherence to ethical AI standards, which contributes to a positive user experience.

Kimi K2 represents a cutting-edge series of open-source large language models utilizing a mixture-of-experts (MoE) architecture, with a staggering 1 trillion parameters in total and 32 billion activated parameters tailored for optimized task execution. Utilizing the Muon optimizer, it has been trained on a substantial dataset of over 15.5 trillion tokens, with its performance enhanced by MuonClip’s attention-logit clamping mechanism, resulting in remarkable capabilities in areas such as advanced knowledge comprehension, logical reasoning, mathematics, programming, and various agentic operations. Moonshot AI offers two distinct versions: Kimi-K2-Base, designed for research-level fine-tuning, and Kimi-K2-Instruct, which is pre-trained for immediate applications in chat and tool interactions, facilitating both customized development and seamless integration of agentic features. Comparative benchmarks indicate that Kimi K2 surpasses other leading open-source models and competes effectively with top proprietary systems, particularly excelling in coding and intricate task analysis. Furthermore, it boasts a generous context length of 128 K tokens, compatibility with tool-calling APIs, and support for industry-standard inference engines, making it a versatile option for various applications. The innovative design and features of Kimi K2 position it as a significant advancement in the field of artificial intelligence language processing.

Kimi K2 Thinking is a sophisticated open-source reasoning model created by Moonshot AI, specifically tailored for intricate, multi-step workflows where it effectively combines chain-of-thought reasoning with tool utilization across numerous sequential tasks. Employing a cutting-edge mixture-of-experts architecture, the model encompasses a staggering total of 1 trillion parameters, although only around 32 billion parameters are utilized during each inference, which enhances efficiency while retaining significant capability. It boasts a context window that can accommodate up to 256,000 tokens, allowing it to process exceptionally long inputs and reasoning sequences without sacrificing coherence. Additionally, it features native INT4 quantization, which significantly cuts down inference latency and memory consumption without compromising performance. Designed with agentic workflows in mind, Kimi K2 Thinking is capable of autonomously invoking external tools, orchestrating sequential logic steps—often involving around 200-300 tool calls in a single chain—and ensuring consistent reasoning throughout the process. Its robust architecture makes it an ideal solution for complex reasoning tasks that require both depth and efficiency.

Kimi K2.5 is a powerful multimodal AI model built to handle complex reasoning, coding, and visual understanding at scale. It supports both text and image or video inputs, enabling developers to build applications that go beyond traditional language-only models. As Kimi’s most advanced model to date, it delivers open-source state-of-the-art performance across agent tasks, software development, and general intelligence benchmarks. The model supports an ultra-long 256K context window, making it ideal for large codebases, long documents, and multi-turn conversations. Kimi K2.5 includes a long-thinking mode that excels at logical reasoning, mathematics, and structured problem solving. It integrates seamlessly with existing workflows through full compatibility with the OpenAI SDK and API format. Developers can use Kimi K2.5 for chat, tool calling, file-based Q&A, and multimodal analysis. Built-in support for streaming, partial mode, and web search expands its flexibility. With predictable pricing and enterprise-ready capabilities, Kimi K2.5 is designed for scalable AI development.

Postgres now features open-source vector similarity search capabilities. This allows for both exact and approximate nearest neighbor searches utilizing L2 distance, inner product, and cosine distance metrics. Additionally, this functionality enhances the database's ability to manage and analyze complex data efficiently.

Kimi is a highly capable assistant equipped with an extensive "memory" that allows her to read lengthy novels of up to 200,000 words and browse the Internet simultaneously. With her ability to comprehend and analyze long documents, Kimi is invaluable for quickly summarizing reports such as financial analyses and research findings, thereby streamlining your reading and organizational tasks. When it comes to studying for exams or delving into new subjects, Kimi can efficiently summarize and clarify complex information from textbooks or academic papers. For those engaged in programming or tech-related tasks, Kimi offers support by reproducing code or suggesting technical solutions based on your input, whether it's code snippets or pseudocode from your documents. Proficient in Chinese and capable of managing multilingual content, Kimi enhances communication and understanding in international settings, making her a versatile tool for global collaboration. Additionally, Kimi Chat can engage you in dynamic conversations or even embody your favorite game characters, providing both entertainment and a way to unwind. Not only does Kimi assist with productivity, but she also brings a fun and interactive element to your daily routine.

Introducing an open-source AI model that can be fine-tuned, distilled, and deployed across various platforms. Our newest instruction-tuned model comes in three sizes: 8B, 70B, and 405B, giving you options to suit different needs. With our open ecosystem, you can expedite your development process using a diverse array of tailored product offerings designed to meet your specific requirements. You have the flexibility to select between real-time inference and batch inference services according to your project's demands. Additionally, you can download model weights to enhance cost efficiency per token while fine-tuning for your application. Improve performance further by utilizing synthetic data and seamlessly deploy your solutions on-premises or in the cloud. Take advantage of Llama system components and expand the model's capabilities through zero-shot tool usage and retrieval-augmented generation (RAG) to foster agentic behaviors. By utilizing 405B high-quality data, you can refine specialized models tailored to distinct use cases, ensuring optimal functionality for your applications. Ultimately, this empowers developers to create innovative solutions that are both efficient and effective.

FLUX.1 represents a revolutionary suite of open-source text-to-image models created by Black Forest Labs, achieving new heights in AI-generated imagery with an impressive 12 billion parameters. This model outperforms established competitors such as Midjourney V6, DALL-E 3, and Stable Diffusion 3 Ultra, providing enhanced image quality, intricate details, high prompt fidelity, and adaptability across a variety of styles and scenes. The FLUX.1 suite is available in three distinct variants: Pro for high-end commercial applications, Dev tailored for non-commercial research with efficiency on par with Pro, and Schnell designed for quick personal and local development initiatives under an Apache 2.0 license. Notably, its pioneering use of flow matching alongside rotary positional embeddings facilitates both effective and high-quality image synthesis. As a result, FLUX.1 represents a significant leap forward in the realm of AI-driven visual creativity, showcasing the potential of advancements in machine learning technology. This model not only elevates the standard for image generation but also empowers creators to explore new artistic possibilities.

The newest version in the Llama series, Llama 3.3, represents a significant advancement in language models aimed at enhancing AI's capabilities in understanding and communication. It boasts improved contextual reasoning, superior language generation, and advanced fine-tuning features aimed at producing exceptionally accurate, human-like responses across a variety of uses. This iteration incorporates a more extensive training dataset, refined algorithms for deeper comprehension, and mitigated biases compared to earlier versions. Llama 3.3 stands out in applications including natural language understanding, creative writing, technical explanations, and multilingual interactions, making it a crucial asset for businesses, developers, and researchers alike. Additionally, its modular architecture facilitates customizable deployment in specific fields, ensuring it remains versatile and high-performing even in large-scale applications. With these enhancements, Llama 3.3 is poised to redefine the standards of AI language models.

The QwQ-32B model, created by Alibaba Cloud's Qwen team, represents a significant advancement in AI reasoning, aimed at improving problem-solving skills. Boasting 32 billion parameters, it rivals leading models such as DeepSeek's R1, which contains 671 billion parameters. This remarkable efficiency stems from its optimized use of parameters, enabling QwQ-32B to tackle complex tasks like mathematical reasoning, programming, and other problem-solving scenarios while consuming fewer resources. It can handle a context length of up to 32,000 tokens, making it adept at managing large volumes of input data. Notably, QwQ-32B is available through Alibaba's Qwen Chat service and is released under the Apache 2.0 license, which fosters collaboration and innovation among AI developers. With its cutting-edge features, QwQ-32B is poised to make a substantial impact in the field of artificial intelligence.

Gemma 3, launched by Google, represents a cutting-edge AI model constructed upon the Gemini 2.0 framework, aimed at delivering superior efficiency and adaptability. This innovative model can operate seamlessly on a single GPU or TPU, which opens up opportunities for a diverse group of developers and researchers. Focusing on enhancing natural language comprehension, generation, and other AI-related functions, Gemma 3 is designed to elevate the capabilities of AI systems. With its scalable and robust features, Gemma 3 aspires to propel the evolution of AI applications in numerous sectors and scenarios, potentially transforming the landscape of technology as we know it.

Qwen3 is a state-of-the-art large language model designed to revolutionize the way we interact with AI. Featuring both thinking and non-thinking modes, Qwen3 allows users to customize its response style, ensuring optimal performance for both complex reasoning tasks and quick inquiries. With the ability to support 119 languages, the model is suitable for international projects. The model's hybrid training approach, which involves over 36 trillion tokens, ensures accuracy across a variety of disciplines, from coding to STEM problems. Its integration with platforms such as Hugging Face, ModelScope, and Kaggle allows for easy adoption in both research and production environments. By enhancing multilingual support and incorporating advanced AI techniques, Qwen3 is designed to push the boundaries of AI-driven applications.

BGE (BAAI General Embedding) serves as a versatile retrieval toolkit aimed at enhancing search capabilities and Retrieval-Augmented Generation (RAG) applications. It encompasses functionalities for inference, evaluation, and fine-tuning of embedding models and rerankers, aiding in the creation of sophisticated information retrieval systems. This toolkit features essential elements such as embedders and rerankers, which are designed to be incorporated into RAG pipelines, significantly improving the relevance and precision of search results. BGE accommodates a variety of retrieval techniques, including dense retrieval, multi-vector retrieval, and sparse retrieval, allowing it to adapt to diverse data types and retrieval contexts. Users can access the models via platforms like Hugging Face, and the toolkit offers a range of tutorials and APIs to help implement and customize their retrieval systems efficiently. By utilizing BGE, developers are empowered to construct robust, high-performing search solutions that meet their unique requirements, ultimately enhancing user experience and satisfaction. Furthermore, the adaptability of BGE ensures it can evolve alongside emerging technologies and methodologies in the data retrieval landscape.

Qwen3-Coder is a versatile coding model that comes in various sizes, prominently featuring the 480B-parameter Mixture-of-Experts version with 35B active parameters, which naturally accommodates 256K-token contexts that can be extended to 1M tokens. This model achieves impressive performance that rivals Claude Sonnet 4, having undergone pre-training on 7.5 trillion tokens, with 70% of that being code, and utilizing synthetic data refined through Qwen2.5-Coder to enhance both coding skills and overall capabilities. Furthermore, the model benefits from post-training techniques that leverage extensive, execution-guided reinforcement learning, which facilitates the generation of diverse test cases across 20,000 parallel environments, thereby excelling in multi-turn software engineering tasks such as SWE-Bench Verified without needing test-time scaling. In addition to the model itself, the open-source Qwen Code CLI, derived from Gemini Code, empowers users to deploy Qwen3-Coder in dynamic workflows with tailored prompts and function calling protocols, while also offering smooth integration with Node.js, OpenAI SDKs, and environment variables. This comprehensive ecosystem supports developers in optimizing their coding projects effectively and efficiently.

Z.ai serves as a versatile, complimentary AI assistant that integrates presentations, writing, and coding into a seamless conversational platform. By harnessing the power of advanced language models, it enables users to create sophisticated slide decks with AI-generated slides, produce high-quality text for various purposes such as emails, reports, and blogs, and even write or troubleshoot intricate code. In addition to content generation, Z.ai excels in conducting thorough research and information retrieval, allowing users to collect data, condense lengthy documents, and break through writer's block, while its coding assistant can clarify code snippets, optimize functions, or generate scripts from the ground up. The user-friendly chat interface eliminates the need for extensive training; you simply communicate your requirements—be it a strategic presentation, marketing content, or a script for data analysis—and receive immediate, contextually pertinent outcomes. With capabilities that extend to multiple languages, including Chinese, as well as native function invocation and support for an extensive 128K token context, Z.ai is equipped to facilitate everything from idea generation to the automation of tedious writing or coding tasks, making it an invaluable tool for professionals across various fields. Its comprehensive approach ensures that users can navigate complex projects with ease and efficiency.

DeepSeek V3.1 stands as a revolutionary open-weight large language model, boasting an impressive 685-billion parameters and an expansive 128,000-token context window, which allows it to analyze extensive documents akin to 400-page books in a single invocation. This model offers integrated functionalities for chatting, reasoning, and code creation, all within a cohesive hybrid architecture that harmonizes these diverse capabilities. Furthermore, V3.1 accommodates multiple tensor formats, granting developers the versatility to enhance performance across various hardware setups. Preliminary benchmark evaluations reveal strong results, including a remarkable 71.6% on the Aider coding benchmark, positioning it competitively with or even superior to systems such as Claude Opus 4, while achieving this at a significantly reduced cost. Released under an open-source license on Hugging Face with little publicity, DeepSeek V3.1 is set to revolutionize access to advanced AI technologies, potentially disrupting the landscape dominated by conventional proprietary models. Its innovative features and cost-effectiveness may attract a wide range of developers eager to leverage cutting-edge AI in their projects.

Hermes 4 represents the cutting-edge advancement in Nous Research's series of neutrally aligned, steerable foundational models, featuring innovative hybrid reasoners that can fluidly transition between creative, expressive outputs and concise, efficient responses tailored to user inquiries. This model is engineered to prioritize user and system commands over any corporate ethical guidelines, resulting in interactions that are more conversational and engaging, avoiding a tone that feels overly authoritative or ingratiating, while fostering opportunities for roleplay and imaginative engagement. By utilizing a specific tag within prompts, users can activate a deeper level of reasoning that is resource-intensive, allowing them to address intricate challenges, all while maintaining efficiency for simpler tasks. With a training dataset 50 times larger than that of its predecessor, Hermes 3, much of which was synthetically produced using Atropos, Hermes 4 exhibits remarkable enhancements in performance. Additionally, this evolution not only improves accuracy but also broadens the range of applications for which the model can be effectively employed.

Devstral Small 2 serves as the streamlined, 24 billion-parameter version of Mistral AI's innovative coding-centric model lineup, released under the flexible Apache 2.0 license to facilitate both local implementations and API interactions. In conjunction with its larger counterpart, Devstral 2, this model introduces "agentic coding" features suitable for environments with limited computational power, boasting a generous 256K-token context window that allows it to comprehend and modify entire codebases effectively. Achieving a score of approximately 68.0% on the standard code-generation evaluation known as SWE-Bench Verified, Devstral Small 2 stands out among open-weight models that are significantly larger. Its compact size and efficient architecture enable it to operate on a single GPU or even in CPU-only configurations, making it an ideal choice for developers, small teams, or enthusiasts lacking access to expansive data-center resources. Furthermore, despite its smaller size, Devstral Small 2 successfully maintains essential functionalities of its larger variants, such as the ability to reason through multiple files and manage dependencies effectively, ensuring that users can still benefit from robust coding assistance. This blend of efficiency and performance makes it a valuable tool in the coding community.

JSON, which stands for JavaScript Object Notation, serves as a compact format for data exchange. Its simplicity makes it accessible for human comprehension and straightforward for machines to interpret and create. Derived from a portion of the JavaScript Programming Language Standard ECMA-262 3rd Edition from December 1999, JSON is a text-based format that remains entirely independent of any specific programming language while employing familiar conventions found in C-family languages such as C, C++, C#, Java, JavaScript, Perl, and Python. This versatility positions JSON as an exceptional choice for data interchange. The structure of JSON is founded on two primary components: 1. A set of name/value pairs, which can be represented in different programming languages as objects, records, structs, dictionaries, hash tables, keyed lists, or associative arrays. 2. An ordered sequence of values, typically manifested in most languages as arrays, vectors, lists, or sequences. These fundamental structures are universally recognized, and nearly all contemporary programming languages incorporate them in some capacity, further enhancing the utility and appeal of JSON as a data format.

A robust platform optimized for training is equipped with NVIDIA® H100 Tensor Core GPUs, offering competitive pricing and personalized support. Designed to handle extensive machine learning workloads, it allows for efficient multihost training across thousands of H100 GPUs interconnected via the latest InfiniBand network, achieving speeds of up to 3.2Tb/s per host. Users benefit from significant cost savings, with at least a 50% reduction in GPU compute expenses compared to leading public cloud services*, and additional savings are available through GPU reservations and bulk purchases. To facilitate a smooth transition, we promise dedicated engineering support that guarantees effective platform integration while optimizing your infrastructure and deploying Kubernetes. Our fully managed Kubernetes service streamlines the deployment, scaling, and management of machine learning frameworks, enabling multi-node GPU training with ease. Additionally, our Marketplace features a variety of machine learning libraries, applications, frameworks, and tools designed to enhance your model training experience. New users can take advantage of a complimentary one-month trial period, ensuring they can explore the platform's capabilities effortlessly. This combination of performance and support makes it an ideal choice for organizations looking to elevate their machine learning initiatives.

Gemma 4 is an advanced AI model developed by Google as part of its Gemini architecture, designed to deliver strong performance while remaining accessible to developers. The model is optimized to run on a single GPU or TPU, allowing more organizations and researchers to experiment with powerful AI technology. Gemma 4 improves natural language understanding and generation, making it suitable for applications such as chatbots, text analysis, and automated content creation. Its architecture enables the model to process complex language patterns while maintaining efficient computational performance. Developers can integrate Gemma 4 into various AI projects that require intelligent text processing or conversational capabilities. The model is designed with scalability in mind, allowing it to support both research experiments and production systems. By offering high-performance AI in a more accessible format, Gemma 4 lowers the barrier for developing sophisticated AI solutions. Its flexibility makes it useful for industries ranging from technology and education to business automation. Researchers can also use the model to explore new AI techniques and improve language processing systems. Overall, Gemma 4 represents a step forward in making powerful AI models easier to deploy and use.

The Gemma family consists of advanced, lightweight models developed using the same innovative research and technology as the Gemini models. These cutting-edge models are equipped with robust security features that promote responsible and trustworthy AI applications, achieved through carefully curated data sets and thorough refinements. Notably, Gemma models excel in their various sizes—2B, 7B, 9B, and 27B—often exceeding the performance of some larger open models. With the introduction of Keras 3.0, users can experience effortless integration with JAX, TensorFlow, and PyTorch, providing flexibility in framework selection based on specific tasks. Designed for peak performance and remarkable efficiency, Gemma 2 is specifically optimized for rapid inference across a range of hardware platforms. Furthermore, the Gemma family includes diverse models that cater to distinct use cases, ensuring they adapt effectively to user requirements. These lightweight language models feature a decoder and have been trained on an extensive array of textual data, programming code, and mathematical concepts, which enhances their versatility and utility in various applications.

Stable Diffusion XL, also known as SDXL, represents the most advanced image generation model, designed specifically to achieve higher levels of photorealism and intricate detail in imagery and composition than earlier versions like SD 2.1. This enhancement allows users to generate images that feature improved facial representations and clearer text, while also enabling the creation of visually appealing artwork with the use of concise prompts. As a result, artists and creators can now express their ideas more effectively and efficiently.

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.

Llama Guard is a collaborative open-source safety model created by Meta AI aimed at improving the security of large language models during interactions with humans. It operates as a filtering mechanism for inputs and outputs, categorizing both prompts and replies based on potential safety risks such as toxicity, hate speech, and false information. With training on a meticulously selected dataset, Llama Guard's performance rivals or surpasses that of existing moderation frameworks, including OpenAI's Moderation API and ToxicChat. This model features an instruction-tuned framework that permits developers to tailor its classification system and output styles to cater to specific applications. As a component of Meta's extensive "Purple Llama" project, it integrates both proactive and reactive security measures to ensure the responsible use of generative AI technologies. The availability of the model weights in the public domain invites additional exploration and modifications to address the continually changing landscape of AI safety concerns, fostering innovation and collaboration in the field. This open-access approach not only enhances the community's ability to experiment but also promotes a shared commitment to ethical AI development.

Z.ai has unveiled its latest flagship model, GLM-4.5, which boasts an impressive 355 billion total parameters (with 32 billion active) and is complemented by the GLM-4.5-Air variant, featuring 106 billion total parameters (12 billion active), designed to integrate sophisticated reasoning, coding, and agent-like functions into a single framework. This model can switch between a "thinking" mode for intricate, multi-step reasoning and tool usage and a "non-thinking" mode that facilitates rapid responses, accommodating a context length of up to 128K tokens and enabling native function invocation. Accessible through the Z.ai chat platform and API, and with open weights available on platforms like HuggingFace and ModelScope, GLM-4.5 is adept at processing a wide range of inputs for tasks such as general problem solving, common-sense reasoning, coding from the ground up or within existing frameworks, as well as managing comprehensive workflows like web browsing and slide generation. The architecture is underpinned by a Mixture-of-Experts design, featuring loss-free balance routing, grouped-query attention mechanisms, and an MTP layer that facilitates speculative decoding, ensuring it meets enterprise-level performance standards while remaining adaptable to various applications. As a result, GLM-4.5 sets a new benchmark for AI capabilities across numerous domains.

gpt-oss-20b is a powerful text-only reasoning model consisting of 20 billion parameters, made available under the Apache 2.0 license and influenced by OpenAI’s gpt-oss usage guidelines, designed to facilitate effortless integration into personalized AI workflows through the Responses API without depending on proprietary systems. It has been specifically trained to excel in instruction following and offers features like adjustable reasoning effort, comprehensive chain-of-thought outputs, and the ability to utilize native tools such as web search and Python execution, resulting in structured and clear responses. Developers are responsible for establishing their own deployment precautions, including input filtering, output monitoring, and adherence to usage policies, to ensure that they align with the protective measures typically found in hosted solutions and to reduce the chance of malicious or unintended actions. Additionally, its open-weight architecture makes it particularly suitable for on-premises or edge deployments, emphasizing the importance of control, customization, and transparency to meet specific user needs. This flexibility allows organizations to tailor the model according to their unique requirements while maintaining a high level of operational integrity.

gpt-oss-120b is a text-only reasoning model with 120 billion parameters, released under the Apache 2.0 license and managed by OpenAI’s usage policy, developed with insights from the open-source community and compatible with the Responses API. It is particularly proficient in following instructions, utilizing tools like web search and Python code execution, and allowing for adjustable reasoning effort, thereby producing comprehensive chain-of-thought and structured outputs that can be integrated into various workflows. While it has been designed to adhere to OpenAI's safety policies, its open-weight characteristics present a risk that skilled individuals might fine-tune it to circumvent these safeguards, necessitating that developers and enterprises apply additional measures to ensure safety comparable to that of hosted models. Evaluations indicate that gpt-oss-120b does not achieve high capability thresholds in areas such as biological, chemical, or cyber domains, even following adversarial fine-tuning. Furthermore, its release is not seen as a significant leap forward in biological capabilities, marking a cautious approach to its deployment. As such, users are encouraged to remain vigilant about the potential implications of its open-weight nature.

Llama (Large Language Model Meta AI) stands as a cutting-edge foundational large language model aimed at helping researchers push the boundaries of their work within this area of artificial intelligence. By providing smaller yet highly effective models like Llama, the research community can benefit even if they lack extensive infrastructure, thus promoting greater accessibility in this dynamic and rapidly evolving domain. Creating smaller foundational models such as Llama is advantageous in the landscape of large language models, as it demands significantly reduced computational power and resources, facilitating the testing of innovative methods, confirming existing research, and investigating new applications. These foundational models leverage extensive unlabeled datasets, making them exceptionally suitable for fine-tuning across a range of tasks. We are offering Llama in multiple sizes (7B, 13B, 33B, and 65B parameters), accompanied by a detailed Llama model card that outlines our development process while adhering to our commitment to Responsible AI principles. By making these resources available, we aim to empower a broader segment of the research community to engage with and contribute to advancements in AI.