Top Timestral Alternatives

TimesFM-3 is an innovative foundation model for time series analysis, distinguished by its capability to deliver highly accurate multivariate forecasts with a single forward pass. With a substantial framework of 330 million parameters, this model has been pre-trained on an extensive dataset comprising both real-world and synthetic time series information, accumulating over 1 trillion time points, which significantly bolsters its effectiveness and zero-shot generalization compared to its predecessors. It excels at simultaneously forecasting multiple coevolving time series while effectively recognizing dependencies that enhance predictive accuracy without necessitating task-specific fine-tuning. Additionally, it is designed to handle various forecasting objectives, including point and quantile predictions, and takes into account both historical covariates and dynamic covariates related to future scenarios, such as planned promotions, holidays, or shifts in weather. Employing a decoder-only transformer architecture, TimesFM-3 adeptly processes sequential data in chunks of 32 time steps, utilizing alternating causal temporal attention and full variate attention to seamlessly weave together patterns across time and interconnected series. As a result, this model serves as a powerful resource for forecasting intricate, time-dependent phenomena across diverse applications, making it a significant advancement in the field of time series forecasting. Its versatility and precision open up new avenues for exploration and application in various domains.

Muse Spark Open is a large-scale open-weight artificial intelligence model developed by Meta for multimodal understanding, software engineering, autonomous agents, and long-horizon reasoning. Released in October 2026, the model provides downloadable weights that allow developers to deploy and customize it outside Meta's hosted AI services. It uses a mixture-of-experts architecture containing 816 billion total parameters, with approximately 42 billion parameters activated during inference. Its one-million-token context window supports processing extensive documents, large software repositories, and lengthy sequences of instructions and interactions. The model is designed for complex coding workflows involving source code analysis, debugging, feature implementation, application development, and software testing. Its multimodal capabilities enable AI agents to interpret visual information and combine that understanding with reasoning and automated computer interactions. Meta demonstrated the model performing local software development tasks and controlling DaVinci Resolve to analyze video footage and prepare edited content. Muse Spark Open supports on-device deployment using two Apple M5 Ultra Mac Studio computers running the MLX framework, with additional compatibility across AMD, Intel, and NVIDIA hardware. Supported inference frameworks include SGLang, vLLM, TensorRT-LLM, Hugging Face Transformers, llama.cpp, and Unsloth, with optimized GPU kernels for improved processing performance. Developers can customize the model using PyTorch and TorchTitan to adapt its behavior to specialized tasks, organizational requirements, and application workflows. Muse Spark Open is distributed under the Muse Model License with limited commercial terms and is available through Hugging Face, Amazon SageMaker AI, Google Cloud Model Garden, Microsoft Foundry Models, and several hosted API providers.

NVIDIA Earth-2 represents a cutting-edge platform that transforms weather and climate forecasting through high-performance, AI-driven technology, integrating various open-source models, libraries, and frameworks to deliver professional-level predictions without relying on traditional supercomputing capabilities. This groundbreaking system is equipped to handle the entire process from collecting raw observational data to producing high-resolution forecasts, including localized storm warnings and medium-range predictions that can extend up to 15 days. By harnessing the power of generative AI architectures, Earth-2 dramatically speeds up computational tasks while either preserving or improving accuracy compared to traditional forecasting methods. The Earth-2 suite includes specialized models like Atlas for multi-variable medium-range forecasting, StormScope for short-term localized weather insights, HealDA for global data assimilation, CorrDiff for enhancing regional resolution, and FourCastNet 3 aimed at effective global predictions. Additionally, this platform not only improves forecasting precision but also democratizes access to sophisticated forecasting tools, enabling a broader audience to benefit from these advancements. As such, Earth-2 marks a significant step forward in making weather forecasting more efficient and accessible to all.

Nixtla is a state-of-the-art platform focused on time-series forecasting and anomaly detection, featuring its groundbreaking model, TimeGPT, which is heralded as the first generative AI foundation model specifically designed for time-series data. Trained on a vast dataset that encompasses over 100 billion data points from various industries, including retail, energy, finance, IoT, healthcare, weather, and web traffic, this model is adept at making accurate zero-shot predictions across a multitude of scenarios. With the help of the Python SDK, users can easily create forecasts or pinpoint anomalies in their datasets using only a few lines of code, even when faced with irregular or sparse time series, eliminating the necessity to build or train models from scratch. Furthermore, TimeGPT is equipped with sophisticated features such as the integration of external influences (like events and pricing), the ability to forecast multiple time series concurrently, the use of custom loss functions, cross-validation capabilities, the provision of prediction intervals, and the option to fine-tune on tailored datasets. This remarkable flexibility positions Nixtla as an essential resource for professionals aiming to elevate their time-series analysis and improve forecasting precision, ultimately facilitating more informed decision-making in their respective fields. Additionally, the platform continuously evolves to incorporate the latest advancements in AI, ensuring that users remain at the forefront of time-series analysis technology.

We are excited to unveil Mistral NeMo, our latest and most sophisticated small model, boasting an impressive 12 billion parameters and a vast context length of 128,000 tokens, all available under the Apache 2.0 license. In collaboration with NVIDIA, Mistral NeMo stands out in its category for its exceptional reasoning capabilities, extensive world knowledge, and coding skills. Its architecture adheres to established industry standards, ensuring it is user-friendly and serves as a smooth transition for those currently using Mistral 7B. To encourage adoption by researchers and businesses alike, we are providing both pre-trained base models and instruction-tuned checkpoints, all under the Apache license. A remarkable feature of Mistral NeMo is its quantization awareness, which enables FP8 inference while maintaining high performance levels. Additionally, the model is well-suited for a range of global applications, showcasing its ability in function calling and offering a significant context window. When benchmarked against Mistral 7B, Mistral NeMo demonstrates a marked improvement in comprehending and executing intricate instructions, highlighting its advanced reasoning abilities and capacity to handle complex multi-turn dialogues. Furthermore, its design not only enhances its performance but also positions it as a formidable option for multi-lingual tasks, ensuring it meets the diverse needs of various use cases while paving the way for future innovations.

Solar Mini is a cutting-edge pre-trained large language model that rivals the capabilities of GPT-3.5 and delivers answers 2.5 times more swiftly, all while keeping its parameter count below 30 billion. In December 2023, it achieved the highest rank on the Hugging Face Open LLM Leaderboard by employing a 32-layer Llama 2 architecture initialized with high-quality Mistral 7B weights, along with a groundbreaking technique called "depth up-scaling" (DUS) that efficiently increases the model's depth without requiring complex modules. After the DUS approach is applied, the model goes through additional pretraining to enhance its performance, and it incorporates instruction tuning designed in a question-and-answer style specifically for Korean, which refines its ability to respond to user queries effectively. Moreover, alignment tuning is implemented to ensure that its outputs are in harmony with human or advanced AI expectations. Solar Mini consistently outperforms competitors such as Llama 2, Mistral 7B, Ko-Alpaca, and KULLM across various benchmarks, proving that innovative architectural approaches can lead to remarkably efficient and powerful AI models. This achievement not only highlights the effectiveness of Solar Mini but also emphasizes the importance of continually evolving strategies in the AI field.

The Mixtral 8x7B model represents a cutting-edge sparse mixture of experts (SMoE) architecture that features open weights and is made available under the Apache 2.0 license. This innovative model outperforms Llama 2 70B across a range of benchmarks, while also achieving inference speeds that are sixfold faster. As the premier open-weight model with a versatile licensing structure, Mixtral stands out for its impressive cost-effectiveness and performance metrics. Furthermore, it competes with and frequently exceeds the capabilities of GPT-3.5 in many established benchmarks, underscoring its importance in the AI landscape. Its unique blend of accessibility, rapid processing, and overall effectiveness positions it as an attractive option for developers in search of top-tier AI solutions. Consequently, the Mixtral model not only enhances the current technological landscape but also paves the way for future advancements in AI development.

Mistral 7B is a cutting-edge language model boasting 7.3 billion parameters, which excels in various benchmarks, even surpassing larger models such as Llama 2 13B. It employs advanced methods like Grouped-Query Attention (GQA) to enhance inference speed and Sliding Window Attention (SWA) to effectively handle extensive sequences. Available under the Apache 2.0 license, Mistral 7B can be deployed across multiple platforms, including local infrastructures and major cloud services. Additionally, a unique variant called Mistral 7B Instruct has demonstrated exceptional abilities in task execution, consistently outperforming rivals like Llama 2 13B Chat in certain applications. This adaptability and performance make Mistral 7B a compelling choice for both developers and researchers seeking efficient solutions. Its innovative features and strong results highlight the model's potential impact on natural language processing projects.

Phonon is an open-weight text-to-speech (TTS) model and inference runtime that caters to developers who wish to integrate speech generation into applications with local and offline audio functionalities. This cutting-edge system converts written text into spoken words and can stream speech in real-time as it is generated. It provides interfaces for programming languages such as Rust and Python, includes a Swift package for iOS and macOS, offers JavaScript support for web browsers, and has libraries available for Android; it can also operate as a self-hosted server compatible with OpenAI's API. The model is designed to support various languages, including English, French, German, Spanish, and Portuguese, depending on the specific variant employed. Phonon is ideal for developers looking to embed speech capabilities into mobile, desktop, web, embedded, and server applications, particularly in situations that prioritize offline use or require enhanced privacy. The English model is comprised of 31 million parameters and is licensed under Apache 2.0, while the language-specific and multilingual models, with 40 million parameters, are available under CC BY-NC 4.0. Furthermore, the runtime's source code is offered under either the Apache 2.0 or MIT licenses, ensuring it is accessible for diverse development requirements. This adaptability and extensive support position Phonon as a highly versatile option for speech synthesis across a variety of platforms, making it a valuable asset for developers in the field.

TabPFN-3.5 represents a cutting-edge foundation model tailored for superior predictions on structured data, proving to be exceptionally valuable for a range of applications including churn analysis, fraud detection, pricing strategies, demand forecasting, and risk assessment, thereby allowing teams to deploy a single model across various use cases. This model efficiently handles data in its native format, managing challenges such as missing values, outliers, categorical variables, multi-table datasets, free text features, and numerous unique identifiers without necessitating any encoding, while also being capable of processing multiple measurements per row. Users benefit from the ability to input raw data directly, eliminating the need for extensive feature engineering or preprocessing, which enables them to receive high-quality, production-ready predictions right after the first prediction call. Significantly, TabPFN-3.5 performs predictions in a single forward pass, achieving an impressive balance between accuracy and speed, and is optimized for rapid inference—a critical aspect for latency-sensitive predictive tasks. Moreover, it can effectively accommodate large datasets of up to one million rows natively and offers an astonishing 20 times faster inference speed compared to earlier versions, marking a significant leap in the domain. This remarkable blend of efficiency, adaptability, and performance establishes TabPFN-3.5 as an invaluable resource for data scientists and organizations aiming to harness structured data to its fullest potential. In addition, the model's user-friendly nature simplifies the workflow, making it accessible for both seasoned experts and those newer to data science.

WeatherNext is a collection of advanced AI-based models created by Google DeepMind and Google Research, aimed at offering state-of-the-art weather forecasting. These innovative models demonstrate superior speed and efficiency compared to traditional physics-based methods, resulting in more reliable forecasts. By enhancing the precision of weather predictions, these advancements have the potential to play a crucial role in disaster preparedness, ultimately helping to save lives in the face of extreme weather events while also improving the reliability of renewable energy systems and supply chains. WeatherNext Graph is particularly notable for providing more accurate and efficient deterministic forecasts than current systems, generating a single forecast for each designated time and location with a 6-hour interval and a 10-day projection. Furthermore, WeatherNext Gen is adept at producing ensemble forecasts that exceed the performance of the leading models, thus granting decision-makers a better grasp of weather uncertainties and the risks linked to extreme weather phenomena. This remarkable enhancement in forecasting capability is set to revolutionize our approach to managing and mitigating the effects of climate variability, ensuring communities are better equipped for future challenges. As a result, the integration of WeatherNext into various sectors could lead to more effective strategies for addressing the complexities of changing weather patterns.

We are excited to unveil an open-source AI model that offers the ability to be fine-tuned, distilled, and deployed across a wide range of platforms. Our latest instruction-tuned model is available in three different sizes: 8B, 70B, and 405B, allowing you to select an option that best fits your unique needs. The open ecosystem we provide accelerates your development journey with a variety of customized product offerings tailored to meet your specific project requirements. You can choose between real-time inference and batch inference services, depending on what your project requires, giving you added flexibility to optimize performance. Furthermore, downloading model weights can significantly enhance cost efficiency per token while you fine-tune the model for your application. To further improve performance, you can leverage synthetic data and seamlessly deploy your solutions either on-premises or in the cloud. By taking advantage of Llama system components, you can also expand the model's capabilities through the use of zero-shot tools and retrieval-augmented generation (RAG), promoting more agentic behaviors in your applications. Utilizing the extensive 405B high-quality data enables you to fine-tune specialized models that cater specifically to various use cases, ensuring that your applications function at their best. In conclusion, this empowers developers to craft innovative solutions that not only meet efficiency standards but also drive effectiveness in their respective domains, leading to a significant impact on the technology landscape.

LFM3 showcases Liquid AI's collection of versatile, open-weight foundational models tailored for a diverse range of applications, spanning text, vision, audio, and multimodal functionalities. This array includes specialized versions for generating text, processing vision-language tasks, handling audio, addressing omni tasks, and facilitating decision-making processes. Developers can choose to download the model weights for customization through fine-tuning and deploy them on local edge devices, or they may opt for Liquid AI’s hosted API for seamless access. The potential uses for these models are extensive, including conversational agents, invoking tools, transcribing and synthesizing speech, interpreting images and videos, extracting and classifying documents, routing information, and enabling robotics capabilities. LFM3 is designed to meet the needs of developers, researchers, and organizations aiming to build customized AI solutions that prioritize adaptable deployment options, including support for on-premises inference. Additionally, these open-weight models are governed by Liquid AI’s LFM Open License, which stipulates the eligibility for commercial use and its specific terms based on the annual revenue of the organization. This level of flexibility empowers users to precisely customize their AI implementations to suit their specific requirements, thereby significantly enhancing the overall functionality and effectiveness of the models. Moreover, with the rapid advancements in AI, having such adaptable tools is crucial for staying competitive in the ever-evolving technological landscape.

Beam marks the launch of Reflection’s first open-weight model, distinguished by its sparse Mixture-of-Experts architecture, which boasts an impressive 501 billion parameters, with 23 billion actively engaged, specifically designed for tasks related to coding, reasoning, and agentic functions. This model's capabilities stem from rigorous pretraining and reinforcement learning, built upon a massive dataset of 23.8 trillion diverse, high-quality tokens obtained from various sources, including the internet, public domains, and proprietary licenses. By focusing on optimizing coding and agentic functionalities, Beam aims to deliver competitive performance in the open-weight space while prioritizing efficient inference computation. It is proficient in managing a diverse range of tasks, such as complex software development, terminal commands, STEM-related activities, web searches, tool application, and general knowledge queries. Utilizing reinforcement learning methods enhances its proficiency in multi-step reasoning, effective tool use, and adaptability to environmental cues. Furthermore, users can customize the model's output by adjusting a reasoning effort parameter, allowing for a tailored balance between efficiency and performance to meet their individual requirements. In essence, Beam is not only a technological advancement but also a versatile tool that empowers users to navigate complex tasks with precision.

TabFM is a cutting-edge foundation model designed for zero-shot learning specifically tailored to manage tabular data, with the goal of simplifying the processes of classification and regression that often demand considerable manual training, hyperparameter tuning, and customized feature engineering. By reframing the difficulties associated with tabular prediction as an in-context learning challenge, TabFM eliminates the necessity of training a distinct supervised model for each dataset; rather, it merges previous training examples with target testing rows into a unified prompt, enabling it to identify the complex relationships that exist between different columns and rows during the inference phase. Since tables are fundamentally two-dimensional and do not depend on a predetermined order, TabFM utilizes a hybrid architecture that combines alternating attention mechanisms for both rows and columns, along with row compression methods, and a dedicated Transformer designed for in-context learning based on these compressed row representations. This advanced structure allows the model to adeptly capture intricate interactions and dependencies among features while ensuring computational efficiency, which is particularly beneficial for dealing with larger datasets. Moreover, this innovative methodology not only boosts performance but also markedly decreases the time and resources generally required for the development of models in tabular data applications, paving the way for more effective analytical solutions. As a result, TabFM represents a significant advancement in the realm of machine learning for tabular data, starting a new era in data analysis.

Liquid AI's LFM2.5 marks a significant evolution in on-device AI foundation models, designed to optimize efficiency and performance for AI inference across edge devices, including smartphones, laptops, vehicles, IoT systems, and various embedded hardware, all while eliminating reliance on cloud computing. This upgraded version builds on the previous LFM2 framework by significantly increasing the scale of pretraining and enhancing the stages of reinforcement learning, leading to a collection of hybrid models that feature approximately 1.2 billion parameters and successfully balance adherence to instructions, reasoning capabilities, and multimodal functions for real-world applications. The LFM2.5 lineup includes various models, such as Base (for fine-tuning and personalization), Instruct (tailored for general-purpose instruction), Japanese-optimized, Vision-Language, and Audio-Language editions, all carefully designed for swift on-device inference, even under strict memory constraints. Additionally, these models are offered as open-weight alternatives, enabling easy deployment through platforms like llama.cpp, MLX, vLLM, and ONNX, which enhances flexibility for developers. With these advancements, LFM2.5 not only solidifies its position as a powerful solution for a wide range of AI-driven tasks but also demonstrates Liquid AI's commitment to pushing the boundaries of what is possible with on-device technology. The combination of scalability and versatility ensures that developers can harness the full potential of AI in practical, everyday scenarios.

Canopy Labs has introduced Orpheus, a groundbreaking collection of advanced speech large language models (LLMs) designed to replicate human-like speech generation. Built on the Llama-3 architecture, these models have been developed using a vast dataset of over 100,000 hours of English speech, enabling them to produce output with natural intonation, emotional nuance, and a rhythmic quality that surpasses current high-end closed-source models. One of the standout features of Orpheus is its zero-shot voice cloning capability, which allows users to replicate voices without needing any prior fine-tuning, alongside user-friendly tags that assist in manipulating emotion and intonation. Engineered for minimal latency, these models achieve around 200ms streaming latency for real-time applications, with potential reductions to approximately 100ms when input streaming is employed. Canopy Labs offers both pre-trained and fine-tuned models featuring 3 billion parameters under the adaptable Apache 2.0 license, and there are plans to develop smaller models with 1 billion, 400 million, and 150 million parameters to accommodate devices with limited processing power. This initiative is anticipated to enhance accessibility and expand the range of applications across diverse platforms and scenarios, making advanced speech generation technology more widely available. As technology continues to evolve, the implications of such advancements could significantly influence fields such as entertainment, education, and customer service.

Recursive Flint is a state-of-the-art language model boasting 40 billion parameters, crafted as a mixture-of-experts particularly for local coding tasks. It utilizes approximately 2.5 billion parameters for every token it processes, allowing for advanced reasoning, multi-turn tool usage, and support for context windows reaching up to 512,000 tokens. Developers have the opportunity to integrate Flint into coding agents, enabling them to inspect codebases, fix bugs, create new features, and formulate tests while running inference directly on their own systems. The model's weights are accessible under the Apache 2.0 license, fostering broader accessibility for users. Flint supports a range of platforms, including llama.cpp, Ollama, Unsloth, LM Studio, vLLM, and SGLang. In addition, Recursive provides rec, an inference engine finely tuned for optimal performance on Apple Silicon M4 Pro and M5 Pro/Max chips. To facilitate the implementation of the model on local machines, detailed setup documentation and practical examples are available for developers. This versatile tool not only aids developers and technical teams in running a code-focused model tailored to their individual system needs but also enhances their overall coding efficiency. As a result, users can significantly boost their development workflows and achieve their programming goals with greater ease and speed than ever before.

Chatterbox is an innovative voice cloning AI model developed by Resemble AI, available as open-source under the MIT license, that enables zero-shot voice cloning using only a five-second audio sample, eliminating the need for lengthy training periods. This model offers advanced speech synthesis with emotional control, allowing users to adjust the expressiveness of the voice from muted to dramatically animated through a simple parameter. Moreover, Chatterbox supports accent adjustments and text-based control, ensuring output that is both high-quality and remarkably human-like. Its ability to provide faster-than-real-time responses makes it an ideal choice for applications that require immediate interaction, such as virtual assistants and immersive media. Tailored for developers, Chatterbox features easy installation through pip and is accompanied by comprehensive documentation. Additionally, it incorporates watermarking technology via Resemble AI’s PerTh (Perceptual Threshold) Watermarker, which subtly embeds information to protect the authenticity of the synthesized audio. This impressive array of features positions Chatterbox as a highly effective tool for crafting diverse and realistic voice applications. As a result, the model not only appeals to developers but also serves as a significant asset in various creative and professional domains. Its focus on user customization and output quality further broadens its potential applications across numerous industries.

Mistral 3 marks the latest development in the realm of open-weight AI models created by Mistral AI, featuring a wide array of options ranging from small, edge-optimized variants to a prominent large-scale multimodal model. Among this selection are three streamlined “Ministral 3” models, equipped with 3 billion, 8 billion, and 14 billion parameters, specifically designed for use on resource-constrained devices like laptops, drones, and various edge devices. In addition, the powerful “Mistral Large 3” serves as a sparse mixture-of-experts model, featuring an impressive total of 675 billion parameters, with 41 billion actively utilized. These models are adept at managing multimodal and multilingual tasks, excelling in areas such as text analysis and image understanding, and have demonstrated remarkable capabilities in responding to general inquiries, handling multilingual conversations, and processing multimodal inputs. Moreover, both the base and instruction-tuned variants are offered under the Apache 2.0 license, which promotes significant customization and integration into a range of enterprise and open-source projects. This approach not only enhances flexibility in usage but also sparks innovation and fosters collaboration among developers and organizations, ultimately driving advancements in AI technology.

Gain a competitive edge and maintain a leading position against rivals by utilizing an innovative automated system that interprets various data sources for actionable insights. Our time series forecasting tool, Dominate, rigorously analyzes economic indicators, global market trends, media dynamics, and other relevant data to enhance supply chain management and predict possible future scenarios. This cutting-edge approach to data analysis has proven effective in some of the most complex environments around the globe. By leveraging artificial intelligence and machine learning, we tap into the connections between extensive data components to significantly influence your outcomes. Our sophisticated multi-step, multi-factor, multi-target autoregressive models can deliver precise predictions for a range of values and adapt them as required. Dominate instills confidence in navigating circumstances to reveal unexpected insights and formulate revolutionary strategies. Additionally, our tensor completion technique proficiently addresses issues with flawed and incomplete data while offering time-series forecasting, alert notifications, and impact evaluations. Ultimately, this powerful capability equips organizations to effectively traverse uncertainty and make empowered, well-informed choices. In today’s fast-paced market, embracing such advanced solutions is essential for sustained success and growth.

Mistral Small 3.1 is an advanced, multimodal, and multilingual AI model that has been made available under the Apache 2.0 license. Building upon the previous Mistral Small 3, this updated version showcases improved text processing abilities and enhanced multimodal understanding, with the capacity to handle an extensive context window of up to 128,000 tokens. It outperforms comparable models like Gemma 3 and GPT-4o Mini, reaching remarkable inference rates of 150 tokens per second. Designed for versatility, Mistral Small 3.1 excels in various applications, including instruction adherence, conversational interaction, visual data interpretation, and executing functions, making it suitable for both commercial and individual AI uses. Its efficient architecture allows it to run smoothly on hardware configurations such as a single RTX 4090 or a Mac with 32GB of RAM, enabling on-device operations. Users have the option to download the model from Hugging Face and explore its features via Mistral AI's developer playground, while it is also embedded in services like Gemini Enterprise Agent Platform and accessible on platforms like NVIDIA NIM. This extensive flexibility empowers developers to utilize its advanced capabilities across a wide range of environments and applications, thereby maximizing its potential impact in the AI landscape. Furthermore, Mistral Small 3.1's innovative design ensures that it remains adaptable to future technological advancements.

Mistral Large 3 is a frontier-scale open AI model built on a sophisticated Mixture-of-Experts framework that unlocks 41B active parameters per step while maintaining a massive 675B total parameter capacity. This architecture lets the model deliver exceptional reasoning, multilingual mastery, and multimodal understanding at a fraction of the compute cost typically associated with models of this scale. Trained entirely from scratch on 3,000 NVIDIA H200 GPUs, it reaches competitive alignment performance with leading closed models, while achieving best-in-class results among permissively licensed alternatives. Mistral Large 3 includes base and instruction editions, supports images natively, and will soon introduce a reasoning-optimized version capable of even deeper thought chains. Its inference stack has been carefully co-designed with NVIDIA, enabling efficient low-precision execution, optimized MoE kernels, speculative decoding, and smooth long-context handling on Blackwell NVL72 systems and enterprise-grade clusters. Through collaborations with vLLM and Red Hat, developers gain an easy path to run Large 3 on single-node 8×A100 or 8×H100 environments with strong throughput and stability. The model is available across Mistral AI Studio, Amazon Bedrock, Azure Foundry, Hugging Face, Fireworks, OpenRouter, Modal, and more, ensuring turnkey access for development teams. Enterprises can go further with Mistral’s custom-training program, tailoring the model to proprietary data, regulatory workflows, or industry-specific tasks. From agentic applications to multilingual customer automation, creative workflows, edge deployment, and advanced tool-use systems, Mistral Large 3 adapts to a wide range of production scenarios. With this release, Mistral positions the 3-series as a complete family—spanning lightweight edge models to frontier-scale MoE intelligence—while remaining fully open, customizable, and performance-optimized across the stack.

PanGu-α is developed with the MindSpore framework and is powered by an impressive configuration of 2048 Ascend 910 AI processors during its training phase. This training leverages a sophisticated parallelism approach through MindSpore Auto-parallel, utilizing five distinct dimensions of parallelism: data parallelism, operation-level model parallelism, pipeline model parallelism, optimizer model parallelism, and rematerialization, to efficiently allocate tasks among the 2048 processors. To enhance the model's generalization capabilities, we compiled an extensive dataset of 1.1TB of high-quality Chinese language information from various domains for pretraining purposes. We rigorously test PanGu-α's generation capabilities across a variety of scenarios, including text summarization, question answering, and dialogue generation. Moreover, we analyze the impact of different model scales on few-shot performance across a broad spectrum of Chinese NLP tasks. Our experimental findings underscore the remarkable performance of PanGu-α, illustrating its proficiency in managing a wide range of tasks, even in few-shot or zero-shot situations, thereby demonstrating its versatility and durability. This thorough assessment not only highlights the strengths of PanGu-α but also emphasizes its promising applications in practical settings. Ultimately, the results suggest that PanGu-α could significantly advance the field of natural language processing.

This repository presents the research preview for LongLLaMA, an innovative large language model capable of handling extensive contexts, reaching up to 256,000 tokens or potentially even more. Built on the OpenLLaMA framework, LongLLaMA has been fine-tuned using the Focused Transformer (FoT) methodology. The foundational code for this model comes from Code Llama. We are excited to introduce a smaller 3B base version of the LongLLaMA model, which is not instruction-tuned, and it will be released under an open license (Apache 2.0). Accompanying this release is inference code that supports longer contexts, available on Hugging Face. The model's weights are designed to effortlessly integrate with existing systems tailored for shorter contexts, particularly those that accommodate up to 2048 tokens. In addition to these features, we provide evaluation results and comparisons to the original OpenLLaMA models, thus offering a thorough insight into LongLLaMA's effectiveness in managing long-context tasks. This advancement marks a significant step forward in the field of language models, enabling more sophisticated applications and research opportunities.

Devstral represents a joint initiative by Mistral AI and All Hands AI, creating an open-source large language model designed explicitly for the field of software engineering. This innovative model exhibits exceptional skill in navigating complex codebases, efficiently managing edits across multiple files, and tackling real-world issues, achieving an impressive 46.8% score on the SWE-Bench Verified benchmark, which positions it ahead of all other open-source models. Built upon the foundation of Mistral-Small-3.1, Devstral features a vast context window that accommodates up to 128,000 tokens. It is optimized for peak performance on advanced hardware configurations, such as Macs with 32GB of RAM or Nvidia RTX 4090 GPUs, and is compatible with several inference frameworks, including vLLM, Transformers, and Ollama. Released under the Apache 2.0 license, Devstral is readily available on various platforms, including Hugging Face, Ollama, Kaggle, Unsloth, and LM Studio, enabling developers to effortlessly incorporate its features into their applications. This model not only boosts efficiency for software engineers but also acts as a crucial tool for anyone engaged in coding tasks, thereby broadening its utility and appeal across the tech community. Furthermore, its open-source nature encourages continuous improvement and collaboration among developers worldwide.

Large language models, which often demand significant computational power and prolonged training periods, have shown remarkable abilities in performing zero- and few-shot learning tasks. The substantial resources required for their creation make it quite difficult for many researchers to replicate these models. Moreover, access to the limited number of models available through APIs is restricted, as users are unable to acquire the full model weights, which hinders academic research. To address these issues, we present Open Pre-trained Transformers (OPT), a series of decoder-only pre-trained transformers that vary in size from 125 million to 175 billion parameters, which we aim to share fully and responsibly with interested researchers. Our research reveals that OPT-175B achieves performance levels comparable to GPT-3, while consuming only one-seventh of the carbon emissions needed for GPT-3's training process. In addition to this, we plan to offer a comprehensive logbook detailing the infrastructural challenges we faced during the project, along with code to aid experimentation with all released models, ensuring that scholars have the necessary resources to further investigate this technology. This initiative not only democratizes access to advanced models but also encourages sustainable practices in the field of artificial intelligence.

Florence-2-large is an advanced vision foundation model developed by Microsoft, aimed at addressing a wide variety of vision and vision-language tasks such as generating captions, recognizing objects, segmenting images, and performing optical character recognition (OCR). It employs a sequence-to-sequence architecture and utilizes the extensive FLD-5B dataset, which contains more than 5 billion annotations along with 126 million images, allowing it to excel in multi-task learning. This model showcases impressive abilities in both zero-shot and fine-tuning contexts, producing outstanding results with minimal training effort. Beyond detailed captioning and object detection, it excels in dense region captioning and can analyze images in conjunction with text prompts to generate relevant responses. Its adaptability enables it to handle a broad spectrum of vision-related challenges through prompt-driven techniques, establishing it as a powerful tool in the domain of AI-powered visual applications. Additionally, users can find this model on Hugging Face, where they can access pre-trained weights that facilitate quick onboarding into image processing tasks. This user-friendly access ensures that both beginners and seasoned professionals can effectively leverage its potential to enhance their projects. As a result, the model not only streamlines the workflow for vision tasks but also encourages innovation within the field by enabling diverse applications.

Aily Labs presents a mobile-focused decision-intelligence platform that empowers businesses with immediate AI-driven insights and predictive analytics across their operations. This cutting-edge solution breaks down data barriers by connecting extensive company data through its Correlator engine, which utilizes advanced AI and machine-learning algorithms to forecast results, simulate scenarios, improve decision-making, and deliver actionable insights to frontline employees. With quick one-day integration and pre-built enterprise workflows, the platform encourages teams to rethink and transform their work methods, ensuring that data is treated as a collective resource while reducing biases in decision-making and enabling quicker, more informed choices. It caters to vital operations like forecasting, simulations, and optimization in supply chains, finance, and other sectors, offering a comprehensive view and decision intelligence right at users' fingertips. Available as an application, this offering is designed to help organizations evolve from passive analysis to proactive decision-making processes, ultimately boosting efficiency and responsiveness. By promoting a culture of data accessibility, Aily Labs helps teams fully leverage their information assets, fostering innovation and driving strategic advancement, which is essential in today's fast-paced business environment. As organizations embrace this platform, they can expect to see significant improvements in collaboration and productivity.

Zyphra is excited to announce the beta launch of Zonos-v0.1, featuring two advanced and real-time text-to-speech models that incorporate high-fidelity voice cloning technology. This release includes a 1.6B transformer model and a 1.6B hybrid model, both distributed under the Apache 2.0 license. Considering the difficulties in measuring audio quality quantitatively, we assert that the quality of output generated by Zonos matches or exceeds that of leading proprietary TTS systems currently on the market. Moreover, we believe that providing access to such high-quality models will significantly enhance progress in TTS research. The model weights for Zonos are readily available on Huggingface, along with sample inference code hosted in our GitHub repository. In addition, Zonos can be accessed through our model playground and API, which offers simple and competitive flat-rate pricing options for users. To showcase Zonos's performance, we have compiled a series of sample comparisons against existing proprietary models that illustrate its exceptional capabilities. This project underscores our dedication to promoting innovation within the text-to-speech technology sector, and we anticipate that it will inspire further advancements in the field.