Alternatives to Yi-Lightning

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.

Qwen2.5-Max is an advanced Mixture-of-Experts (MoE) model created by the Qwen team, which has been pretrained on an extensive dataset of over 20 trillion tokens and subsequently enhanced through methods like Supervised Fine-Tuning (SFT) and Reinforcement Learning from Human Feedback (RLHF). Its performance in evaluations surpasses that of models such as DeepSeek V3 across various benchmarks, including Arena-Hard, LiveBench, LiveCodeBench, and GPQA-Diamond, while also achieving strong results in other tests like MMLU-Pro. This model is available through an API on Alibaba Cloud, allowing users to easily integrate it into their applications, and it can also be interacted with on Qwen Chat for a hands-on experience. With its superior capabilities, Qwen2.5-Max represents a significant advancement in AI model technology.

DeepSeek-V2 is a cutting-edge Mixture-of-Experts (MoE) language model developed by DeepSeek-AI, noted for its cost-effective training and high-efficiency inference features. It boasts an impressive total of 236 billion parameters, with only 21 billion active for each token, and is capable of handling a context length of up to 128K tokens. The model utilizes advanced architectures such as Multi-head Latent Attention (MLA) to optimize inference by minimizing the Key-Value (KV) cache and DeepSeekMoE to enable economical training through sparse computations. Compared to its predecessor, DeepSeek 67B, this model shows remarkable improvements, achieving a 42.5% reduction in training expenses, a 93.3% decrease in KV cache size, and a 5.76-fold increase in generation throughput. Trained on an extensive corpus of 8.1 trillion tokens, DeepSeek-V2 demonstrates exceptional capabilities in language comprehension, programming, and reasoning tasks, positioning it as one of the leading open-source models available today. Its innovative approach not only elevates its performance but also sets new benchmarks within the field of artificial intelligence.

We are thrilled to present DBRX, a versatile open LLM developed by Databricks. This innovative model achieves unprecedented performance on a variety of standard benchmarks, setting a new benchmark for existing open LLMs. Additionally, it equips both the open-source community and enterprises crafting their own LLMs with features that were once exclusive to proprietary model APIs; our evaluations indicate that it outperforms GPT-3.5 and competes effectively with Gemini 1.0 Pro. Notably, it excels as a code model, outperforming specialized counterparts like CodeLLaMA-70B in programming tasks, while also demonstrating its prowess as a general-purpose LLM. The remarkable quality of DBRX is complemented by significant enhancements in both training and inference efficiency. Thanks to its advanced fine-grained mixture-of-experts (MoE) architecture, DBRX elevates the efficiency of open models to new heights. In terms of inference speed, it can be twice as fast as LLaMA2-70B, and its total and active parameter counts are approximately 40% of those in Grok-1, showcasing its compact design without compromising capability. This combination of speed and size makes DBRX a game-changer in the landscape of open AI models.

GigaChat 3 Ultra redefines open-source scale by delivering a 702B-parameter frontier model purpose-built for Russian and multilingual understanding. Designed with a modern MoE architecture, it achieves the reasoning strength of giant dense models while using only a fraction of active parameters per generation step. Its massive 14T-token training corpus includes natural human text, curated multilingual sources, extensive STEM materials, and billions of high-quality synthetic examples crafted to boost logic, math, and programming skills. This model is not a derivative or retrained foreign LLM—it is a ground-up build engineered to capture cultural nuance, linguistic accuracy, and reliable long-context performance. GigaChat 3 Ultra integrates seamlessly with open-source tooling like vLLM, sglang, DeepSeek-class architectures, and HuggingFace-based training stacks. It supports advanced capabilities including a code interpreter, improved chat template, memory system, contextual search reformulation, and 128K context windows. Benchmarking shows clear improvements over previous GigaChat generations and competitive results against global leaders in coding, reasoning, and cross-domain tasks. Overall, GigaChat 3 Ultra empowers teams to explore frontier-scale AI without sacrificing transparency, customizability, or ecosystem compatibility.

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.

Yi-Large is an innovative proprietary large language model created by 01.AI, featuring an impressive context length of 32k and a cost structure of $2 for each million tokens for both inputs and outputs. Renowned for its superior natural language processing abilities, common-sense reasoning, and support for multiple languages, it competes effectively with top models such as GPT-4 and Claude3 across various evaluations. This model is particularly adept at handling tasks that involve intricate inference, accurate prediction, and comprehensive language comprehension, making it ideal for applications such as knowledge retrieval, data categorization, and the development of conversational chatbots that mimic human interaction. Built on a decoder-only transformer architecture, Yi-Large incorporates advanced features like pre-normalization and Group Query Attention, and it has been trained on an extensive, high-quality multilingual dataset to enhance its performance. The model's flexibility and economical pricing position it as a formidable player in the artificial intelligence landscape, especially for businesses looking to implement AI technologies on a global scale. Additionally, its ability to adapt to a wide range of use cases underscores its potential to revolutionize how organizations leverage language models for various needs.

Introducing the next iteration of our open-source large language model, this version features model weights along with initial code for the pretrained and fine-tuned Llama language models, which span from 7 billion to 70 billion parameters. The Llama 2 pretrained models have been developed using an impressive 2 trillion tokens and offer double the context length compared to their predecessor, Llama 1. Furthermore, the fine-tuned models have been enhanced through the analysis of over 1 million human annotations. Llama 2 demonstrates superior performance against various other open-source language models across multiple external benchmarks, excelling in areas such as reasoning, coding capabilities, proficiency, and knowledge assessments. For its training, Llama 2 utilized publicly accessible online data sources, while the fine-tuned variant, Llama-2-chat, incorporates publicly available instruction datasets along with the aforementioned extensive human annotations. Our initiative enjoys strong support from a diverse array of global stakeholders who are enthusiastic about our open approach to AI, including companies that have provided valuable early feedback and are eager to collaborate using Llama 2. The excitement surrounding Llama 2 signifies a pivotal shift in how AI can be developed and utilized collectively.

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.

Qwen3.5 represents a major advancement in open-weight multimodal AI models, engineered to function as a native vision-language agent system. Its flagship model, Qwen3.5-397B-A17B, leverages a hybrid architecture that fuses Gated DeltaNet linear attention with a high-sparsity mixture-of-experts framework, allowing only 17 billion parameters to activate during inference for improved speed and cost efficiency. Despite its sparse activation, the full 397-billion-parameter model achieves competitive performance across reasoning, coding, multilingual benchmarks, and complex agent evaluations. The hosted Qwen3.5-Plus version supports a one-million-token context window and includes built-in tool use for search, code interpretation, and adaptive reasoning. The model significantly expands multilingual coverage to 201 languages and dialects while improving encoding efficiency with a larger vocabulary. Native multimodal training enables strong performance in image understanding, video processing, document analysis, and spatial reasoning tasks. Its infrastructure includes FP8 precision pipelines and heterogeneous parallelism to boost throughput and reduce memory consumption. Reinforcement learning at scale enhances multi-step planning and general agent behavior across text and multimodal environments. Overall, Qwen3.5 positions itself as a high-efficiency foundation for autonomous digital agents capable of reasoning, searching, coding, and interacting with complex environments.

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.

The Xiaomi MiMo API open platform serves as a developer-centric interface that allows for the integration and access of Xiaomi’s MiMo AI model family, which includes various reasoning and language models like MiMo-V2-Flash, enabling the creation of applications and services via standardized APIs and cloud endpoints. This platform empowers developers to incorporate AI-driven functionalities such as conversational agents, reasoning processes, code assistance, and search-enhanced tasks without the need to handle the complexities of model infrastructure. It features RESTful API access complete with authentication, request signing, and well-structured responses, allowing software to send user queries and receive generated text or processed results in a programmatic manner. The platform also supports essential operations including text generation, prompt management, and model inference, facilitating seamless interactions with MiMo models. Furthermore, it provides comprehensive documentation and onboarding resources, enabling teams to effectively integrate the latest open-source large language models from Xiaomi, which utilize innovative Mixture-of-Experts (MoE) architectures to enhance performance and efficiency. Overall, this open platform significantly lowers the barriers for developers looking to harness advanced AI capabilities in their projects.

Tülu 3 is a cutting-edge language model created by the Allen Institute for AI (Ai2) that aims to improve proficiency in fields like knowledge, reasoning, mathematics, coding, and safety. It is based on the Llama 3 Base and undergoes a detailed four-stage post-training regimen: careful prompt curation and synthesis, supervised fine-tuning on a wide array of prompts and completions, preference tuning utilizing both off- and on-policy data, and a unique reinforcement learning strategy that enhances targeted skills through measurable rewards. Notably, this open-source model sets itself apart by ensuring complete transparency, offering access to its training data, code, and evaluation tools, thus bridging the performance divide between open and proprietary fine-tuning techniques. Performance assessments reveal that Tülu 3 surpasses other models with comparable sizes, like Llama 3.1-Instruct and Qwen2.5-Instruct, across an array of benchmarks, highlighting its effectiveness. The continuous development of Tülu 3 signifies the commitment to advancing AI capabilities while promoting an open and accessible approach to technology.

Mistral Large 3 pushes open-source AI into frontier territory with a massive sparse MoE architecture that activates 41B parameters per token while maintaining a highly efficient 675B total parameter design. It sets a new performance standard by combining long-context reasoning, multilingual fluency across 40+ languages, and robust multimodal comprehension within a single unified model. Trained end-to-end on thousands of NVIDIA H200 GPUs, it reaches parity with top closed-source instruction models while remaining fully accessible under the Apache 2.0 license. Developers benefit from optimized deployments through partnerships with NVIDIA, Red Hat, and vLLM, enabling smooth inference on A100, H100, and Blackwell-class systems. The model ships in both base and instruct variants, with a reasoning-enhanced version on the way for even deeper analytical capabilities. Beyond general intelligence, Mistral Large 3 is engineered for enterprise customization, allowing organizations to refine the model on internal datasets or domain-specific tasks. Its efficient token generation and powerful multimodal stack make it ideal for coding, document analysis, knowledge workflows, agentic systems, and multilingual communications. With Mistral Large 3, organizations can finally deploy frontier-class intelligence with full transparency, flexibility, and control.

Qwen3.5-35B-A3B is a member of the Qwen3.5 "Medium" model series, meticulously crafted as an effective multimodal foundation model that strikes a balance between robust reasoning capabilities and practical application needs. Utilizing a Mixture-of-Experts (MoE) architecture, it boasts a total of 35 billion parameters, yet activates only around 3 billion for each token, enabling it to achieve performance levels similar to much larger models while significantly cutting down on computational expenses. The model employs a hybrid attention mechanism that merges linear attention with traditional attention layers, which enhances its ability to handle extensive context and boosts scalability for intricate tasks. As an inherently vision-language model, it processes both textual and visual data, catering to a variety of applications, including multimodal reasoning, programming, and automated workflows. Furthermore, it is engineered to operate as a versatile "AI agent," proficient in planning, utilizing tools, and systematically solving problems, extending its functionality beyond mere conversational interactions. This capability positions it as a valuable asset across diverse domains, where advanced AI-driven solutions are increasingly required.

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.

Qwen2 represents a collection of extensive language models crafted by the Qwen team at Alibaba Cloud. This series encompasses a variety of models, including base and instruction-tuned versions, with parameters varying from 0.5 billion to an impressive 72 billion, showcasing both dense configurations and a Mixture-of-Experts approach. The Qwen2 series aims to outperform many earlier open-weight models, including its predecessor Qwen1.5, while also striving to hold its own against proprietary models across numerous benchmarks in areas such as language comprehension, generation, multilingual functionality, programming, mathematics, and logical reasoning. Furthermore, this innovative series is poised to make a significant impact in the field of artificial intelligence, offering enhanced capabilities for a diverse range of applications.

CodeQwen serves as the coding counterpart to Qwen, which is a series of large language models created by the Qwen team at Alibaba Cloud. Built on a transformer architecture that functions solely as a decoder, this model has undergone extensive pre-training using a vast dataset of code. It showcases robust code generation abilities and demonstrates impressive results across various benchmarking tests. With the capacity to comprehend and generate long contexts of up to 64,000 tokens, CodeQwen accommodates 92 programming languages and excels in tasks such as text-to-SQL queries and debugging. Engaging with CodeQwen is straightforward—you can initiate a conversation with just a few lines of code utilizing transformers. The foundation of this interaction relies on constructing the tokenizer and model using pre-existing methods, employing the generate function to facilitate dialogue guided by the chat template provided by the tokenizer. In alignment with our established practices, we implement the ChatML template tailored for chat models. This model adeptly completes code snippets based on the prompts it receives, delivering responses without the need for any further formatting adjustments, thereby enhancing the user experience. The seamless integration of these elements underscores the efficiency and versatility of CodeQwen in handling diverse coding tasks.

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.

Qwen3.5-Plus is an advanced multimodal foundation model engineered to deliver efficient large-context reasoning across text, image, and video inputs. Powered by a hybrid architecture that merges linear attention mechanisms with a sparse mixture-of-experts framework, the model achieves state-of-the-art performance while reducing computational overhead. It supports deep thinking mode, enabling extended reasoning chains of up to 80K tokens and total context windows of up to 1 million tokens. Developers can leverage features such as structured output generation, function calling, web search, and integrated code interpretation to build intelligent agent workflows. The model is optimized for high throughput, supporting large token-per-minute limits and robust rate limits for enterprise-scale applications. Qwen3.5-Plus also includes explicit caching options to reduce costs during repeated inference tasks. With tiered pricing based on input and output tokens, organizations can scale usage predictably. OpenAI-compatible API endpoints make integration straightforward across existing AI stacks and developer tools. Designed for demanding applications, Qwen3.5-Plus excels in long-document analysis, multimodal reasoning, and advanced AI agent development.

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

This repository showcases the research preview of LongLLaMA, an advanced large language model that can manage extensive contexts of up to 256,000 tokens or potentially more. LongLLaMA is developed on the OpenLLaMA framework and has been fine-tuned utilizing the Focused Transformer (FoT) technique. The underlying code for LongLLaMA is derived from Code Llama. We are releasing a smaller 3B base variant of the LongLLaMA model, which is not instruction-tuned, under an open license (Apache 2.0), along with inference code that accommodates longer contexts available on Hugging Face. This model's weights can seamlessly replace LLaMA in existing systems designed for shorter contexts, specifically those handling up to 2048 tokens. Furthermore, we include evaluation results along with comparisons to the original OpenLLaMA models, thereby providing a comprehensive overview of LongLLaMA's capabilities in the realm of long-context processing.

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

The Mixtral 8x22B represents our newest open model, establishing a new benchmark for both performance and efficiency in the AI sector. This sparse Mixture-of-Experts (SMoE) model activates only 39B parameters from a total of 141B, ensuring exceptional cost efficiency relative to its scale. Additionally, it demonstrates fluency in multiple languages, including English, French, Italian, German, and Spanish, while also possessing robust skills in mathematics and coding. With its native function calling capability, combined with the constrained output mode utilized on la Plateforme, it facilitates the development of applications and the modernization of technology stacks on a large scale. The model's context window can handle up to 64K tokens, enabling accurate information retrieval from extensive documents. We prioritize creating models that maximize cost efficiency for their sizes, thereby offering superior performance-to-cost ratios compared to others in the community. The Mixtral 8x22B serves as a seamless extension of our open model lineage, and its sparse activation patterns contribute to its speed, making it quicker than any comparable dense 70B model on the market. Furthermore, its innovative design positions it as a leading choice for developers seeking high-performance solutions.

MiniMax M2 is an open-source foundational model tailored for agent-driven applications and coding tasks, achieving an innovative equilibrium of efficiency, velocity, and affordability. It shines in comprehensive development environments, adeptly managing programming tasks, invoking tools, and executing intricate, multi-step processes, complete with features like Python integration, while offering impressive inference speeds of approximately 100 tokens per second and competitive API pricing at around 8% of similar proprietary models. The model includes a "Lightning Mode" designed for rapid, streamlined agent operations, alongside a "Pro Mode" aimed at thorough full-stack development, report creation, and the orchestration of web-based tools; its weights are entirely open source, allowing for local deployment via vLLM or SGLang. MiniMax M2 stands out as a model ready for production use, empowering agents to autonomously perform tasks such as data analysis, software development, tool orchestration, and implementing large-scale, multi-step logic across real organizational contexts. With its advanced capabilities, this model is poised to revolutionize the way developers approach complex programming challenges.

MiMo-V2-Flash is a large language model created by Xiaomi that utilizes a Mixture-of-Experts (MoE) framework, combining remarkable performance with efficient inference capabilities. With a total of 309 billion parameters, it activates just 15 billion parameters during each inference, allowing it to effectively balance reasoning quality and computational efficiency. This model is well-suited for handling lengthy contexts, making it ideal for tasks such as long-document comprehension, code generation, and multi-step workflows. Its hybrid attention mechanism integrates both sliding-window and global attention layers, which helps to minimize memory consumption while preserving the ability to understand long-range dependencies. Additionally, the Multi-Token Prediction (MTP) design enhances inference speed by enabling the simultaneous processing of batches of tokens. MiMo-V2-Flash boasts impressive generation rates of up to approximately 150 tokens per second and is specifically optimized for applications that demand continuous reasoning and multi-turn interactions. The innovative architecture of this model reflects a significant advancement in the field of language processing.

In honor of Cleopatra, whose magnificent fate concluded amidst the tragic incident involving a snake, we are excited to introduce Codestral Mamba, a Mamba2 language model specifically designed for code generation and released under an Apache 2.0 license. Codestral Mamba represents a significant advancement in our ongoing initiative to explore and develop innovative architectures. It is freely accessible for use, modification, and distribution, and we aspire for it to unlock new avenues in architectural research. The Mamba models are distinguished by their linear time inference capabilities and their theoretical potential to handle sequences of infinite length. This feature enables users to interact with the model effectively, providing rapid responses regardless of input size. Such efficiency is particularly advantageous for enhancing code productivity; therefore, we have equipped this model with sophisticated coding and reasoning skills, allowing it to perform competitively with state-of-the-art transformer-based models. As we continue to innovate, we believe Codestral Mamba will inspire further advancements in the coding community.

StarCoder and StarCoderBase represent advanced Large Language Models specifically designed for code, developed using openly licensed data from GitHub, which encompasses over 80 programming languages, Git commits, GitHub issues, and Jupyter notebooks. In a manner akin to LLaMA, we constructed a model with approximately 15 billion parameters trained on a staggering 1 trillion tokens. Furthermore, we tailored the StarCoderBase model with 35 billion Python tokens, leading to the creation of what we now refer to as StarCoder. Our evaluations indicated that StarCoderBase surpasses other existing open Code LLMs when tested against popular programming benchmarks and performs on par with or even exceeds proprietary models like code-cushman-001 from OpenAI, the original Codex model that fueled early iterations of GitHub Copilot. With an impressive context length exceeding 8,000 tokens, the StarCoder models possess the capability to handle more information than any other open LLM, thus paving the way for a variety of innovative applications. This versatility is highlighted by our ability to prompt the StarCoder models through a sequence of dialogues, effectively transforming them into dynamic technical assistants that can provide support in diverse programming tasks.

Qwen3.6 is an advanced AI model from Alibaba that builds on previous Qwen releases with a focus on real-world utility and performance. It is designed as a multimodal large language model capable of understanding and generating text while also processing visual and structured data. The model is optimized for coding tasks, enabling developers to handle complex, repository-level programming workflows. Qwen3.6 uses a mixture-of-experts (MoE) architecture, which activates only a portion of its parameters during inference to improve efficiency. This design allows it to deliver strong performance while reducing computational costs. It is available in both proprietary and open-weight versions, giving developers flexibility in deployment. The model supports integration into enterprise systems and cloud platforms, particularly within Alibaba’s ecosystem. Qwen3.6 also introduces stronger agentic capabilities, allowing it to perform multi-step reasoning and more autonomous task execution. It is designed to handle complex workflows, including engineering, analysis, and decision-making tasks. The model emphasizes stability and responsiveness based on developer feedback. Overall, Qwen3.6 provides a scalable and efficient AI solution for coding, automation, and multimodal applications.

Reka Flash 3 is a cutting-edge multimodal AI model with 21 billion parameters, crafted by Reka AI to perform exceptionally well in tasks such as general conversation, coding, following instructions, and executing functions. This model adeptly handles and analyzes a myriad of inputs, including text, images, video, and audio, providing a versatile and compact solution for a wide range of applications. Built from the ground up, Reka Flash 3 was trained on a rich array of datasets, encompassing both publicly available and synthetic information, and it underwent a meticulous instruction tuning process with high-quality selected data to fine-tune its capabilities. The final phase of its training involved employing reinforcement learning techniques, specifically using the REINFORCE Leave One-Out (RLOO) method, which combined both model-based and rule-based rewards to significantly improve its reasoning skills. With an impressive context length of 32,000 tokens, Reka Flash 3 competes effectively with proprietary models like OpenAI's o1-mini, making it an excellent choice for applications requiring low latency or on-device processing. The model operates at full precision with a memory requirement of 39GB (fp16), although it can be efficiently reduced to just 11GB through the use of 4-bit quantization, demonstrating its adaptability for various deployment scenarios. Overall, Reka Flash 3 represents a significant advancement in multimodal AI technology, capable of meeting diverse user needs across multiple platforms.

Mistral Small 4 is a next-generation open-source AI model created by Mistral AI to deliver powerful reasoning, coding, and multimodal capabilities within a single unified architecture. The model merges features from several specialized systems, including Magistral for advanced reasoning, Pixtral for multimodal processing, and Devstral for agentic software development tasks. It supports both text and image inputs, enabling applications such as conversational AI, document analysis, and visual data interpretation. The model is built using a mixture-of-experts design with 128 experts, allowing efficient scaling while maintaining strong performance across diverse tasks. Users can adjust the model’s reasoning behavior through a configurable parameter that toggles between lightweight responses and deeper analytical processing. Mistral Small 4 also provides a large context window that enables it to handle long conversations, detailed documents, and complex reasoning chains. Compared with earlier versions, the model offers improved performance, reduced latency, and higher throughput for real-time applications. Developers can integrate it with popular machine learning frameworks such as Transformers, vLLM, and llama.cpp. The model’s open-source Apache 2.0 license allows organizations to fine-tune and customize it for specialized use cases. By combining efficiency, flexibility, and multimodal intelligence, Mistral Small 4 provides a versatile foundation for building advanced AI-powered applications.

Grok 4.1 Fast represents xAI’s leap forward in building highly capable agents that rely heavily on tool calling, long-context reasoning, and real-time information retrieval. It supports a robust 2-million-token window, enabling long-form planning, deep research, and multi-step workflows without degradation. Through extensive RL training and exposure to diverse tool ecosystems, the model performs exceptionally well on demanding benchmarks like τ²-bench Telecom. When paired with the Agent Tools API, it can autonomously browse the web, search X posts, execute Python code, and retrieve documents, eliminating the need for developers to manage external infrastructure. It is engineered to maintain intelligence across multi-turn conversations, making it ideal for enterprise tasks that require continuous context. Its benchmark accuracy on tool-calling and function-calling tasks clearly surpasses competing models in speed, cost, and reliability. Developers can leverage these strengths to build agents that automate customer support, perform real-time analysis, and execute complex domain-specific tasks. With its performance, low pricing, and availability on platforms like OpenRouter, Grok 4.1 Fast stands out as a production-ready solution for next-generation AI systems.

OpenAI’s GPT-5 mini is a cost-efficient, faster version of the flagship GPT-5 model, designed to handle well-defined tasks and precise inputs with high reasoning capabilities. Supporting text and image inputs, GPT-5 mini can process and generate large amounts of content thanks to its extensive 400,000-token context window and a maximum output of 128,000 tokens. This model is optimized for speed, making it ideal for developers and businesses needing quick turnaround times on natural language processing tasks while maintaining accuracy. The pricing model offers significant savings, charging $0.25 per million input tokens and $2 per million output tokens, compared to the higher costs of the full GPT-5. It supports many advanced API features such as streaming responses, function calling, and fine-tuning, while excluding audio input and image generation capabilities. GPT-5 mini is compatible with a broad range of API endpoints including chat completions, real-time responses, and embeddings, making it highly flexible. Rate limits vary by usage tier, supporting from hundreds to tens of thousands of requests per minute, ensuring reliability for different scale needs. This model strikes a balance between performance and cost, suitable for applications requiring fast, high-quality AI interaction without extensive resource use.

CodeGemma represents an impressive suite of efficient and versatile models capable of tackling numerous coding challenges, including middle code completion, code generation, natural language processing, mathematical reasoning, and following instructions. It features three distinct model types: a 7B pre-trained version designed for code completion and generation based on existing code snippets, a 7B variant fine-tuned for translating natural language queries into code and adhering to instructions, and an advanced 2B pre-trained model that offers code completion speeds up to twice as fast. Whether you're completing lines, developing functions, or crafting entire segments of code, CodeGemma supports your efforts, whether you're working in a local environment or leveraging Google Cloud capabilities. With training on an extensive dataset comprising 500 billion tokens predominantly in English, sourced from web content, mathematics, and programming languages, CodeGemma not only enhances the syntactical accuracy of generated code but also ensures its semantic relevance, thereby minimizing mistakes and streamlining the debugging process. This powerful tool continues to evolve, making coding more accessible and efficient for developers everywhere.

OpenAI’s GPT-5 represents the cutting edge in AI language models, designed to be smarter, faster, and more reliable across diverse applications such as legal analysis, scientific research, and financial modeling. This flagship model incorporates built-in “thinking” to deliver accurate, professional, and nuanced responses that help users solve complex problems. With a massive context window and high token output limits, GPT-5 supports extensive conversations and intricate coding tasks with minimal prompting. It introduces advanced features like the verbosity parameter, enabling users to control the detail and tone of generated content. GPT-5 also integrates seamlessly with enterprise data sources like Google Drive and SharePoint, enhancing response relevance with company-specific knowledge while ensuring data privacy. The model’s improved personality and steerability make it adaptable for a wide range of business needs. Available in ChatGPT and API platforms, GPT-5 brings expert intelligence to every user, from casual individuals to large organizations. Its release marks a major step forward in AI-assisted productivity and collaboration.

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

We are excited to present MPT-7B, the newest addition to the MosaicML Foundation Series. This transformer model has been meticulously trained from the ground up using 1 trillion tokens of diverse text and code. It is open-source and ready for commercial applications, delivering performance on par with LLaMA-7B. The training process took 9.5 days on the MosaicML platform, requiring no human input and incurring an approximate cost of $200,000. With MPT-7B, you can now train, fine-tune, and launch your own customized MPT models, whether you choose to begin with one of our provided checkpoints or start anew. To provide additional options, we are also introducing three fine-tuned variants alongside the base MPT-7B: MPT-7B-Instruct, MPT-7B-Chat, and MPT-7B-StoryWriter-65k+, the latter boasting an impressive context length of 65,000 tokens, allowing for extensive content generation. These advancements open up new possibilities for developers and researchers looking to leverage the power of transformer models in their projects.

OLMo 2 represents a collection of completely open language models created by the Allen Institute for AI (AI2), aimed at giving researchers and developers clear access to training datasets, open-source code, reproducible training methodologies, and thorough assessments. These models are trained on an impressive volume of up to 5 trillion tokens and compete effectively with top open-weight models like Llama 3.1, particularly in English academic evaluations. A key focus of OLMo 2 is on ensuring training stability, employing strategies to mitigate loss spikes during extended training periods, and applying staged training interventions in the later stages of pretraining to mitigate weaknesses in capabilities. Additionally, the models leverage cutting-edge post-training techniques derived from AI2's Tülu 3, leading to the development of OLMo 2-Instruct models. To facilitate ongoing enhancements throughout the development process, an actionable evaluation framework known as the Open Language Modeling Evaluation System (OLMES) was created, which includes 20 benchmarks that evaluate essential capabilities. This comprehensive approach not only fosters transparency but also encourages continuous improvement in language model performance.

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

Amazon Nova Micro is an advanced text-only AI model optimized for rapid language processing at a very low cost. With capabilities in reasoning, translation, and code completion, it offers over 200 tokens per second in response generation, making it suitable for fast-paced, real-time applications. Nova Micro supports fine-tuning with text inputs, and its efficiency in understanding and generating text makes it a cost-effective solution for AI-driven applications requiring high performance and quick outputs.

Grok 3 Think, the newest version of xAI's AI model, aims to significantly improve reasoning skills through sophisticated reinforcement learning techniques. It possesses the ability to analyze intricate issues for durations ranging from mere seconds to several minutes, enhancing its responses by revisiting previous steps, considering different options, and fine-tuning its strategies. This model has been developed on an unparalleled scale, showcasing outstanding proficiency in various tasks, including mathematics, programming, and general knowledge, and achieving notable success in competitions such as the American Invitational Mathematics Examination. Additionally, Grok 3 Think not only yields precise answers but also promotes transparency by enabling users to delve into the rationale behind its conclusions, thereby establishing a new benchmark for artificial intelligence in problem-solving. Its unique approach to transparency and reasoning offers users greater trust and understanding of AI decision-making processes.

Sky-T1-32B-Preview is an innovative open-source reasoning model crafted by the NovaSky team at UC Berkeley's Sky Computing Lab. It delivers performance comparable to proprietary models such as o1-preview on various reasoning and coding assessments, while being developed at a cost of less than $450, highlighting the potential for budget-friendly, advanced reasoning abilities. Fine-tuned from Qwen2.5-32B-Instruct, the model utilized a meticulously curated dataset comprising 17,000 examples spanning multiple fields, such as mathematics and programming. The entire training process was completed in just 19 hours using eight H100 GPUs with DeepSpeed Zero-3 offloading technology. Every component of this initiative—including the data, code, and model weights—is entirely open-source, allowing both academic and open-source communities to not only replicate but also improve upon the model's capabilities. This accessibility fosters collaboration and innovation in the realm of artificial intelligence research and development.

Nemotron 3 Nano is a small yet powerful large language model from NVIDIA's Nemotron 3 series, specifically crafted for effective agentic reasoning, interactive dialogue, and programming assignments. Its innovative Mixture-of-Experts Mamba-Transformer framework selectively activates a limited set of parameters for each token, ensuring rapid inference times without sacrificing accuracy or reasoning capabilities. With roughly 31.6 billion parameters in total, including about 3.2 billion active ones (or 3.6 billion when factoring in embeddings), it surpasses the performance of the previous Nemotron 2 Nano model while requiring less computational effort for each forward pass. The model is equipped to manage long-context processing of up to one million tokens, which allows it to efficiently process extensive documents, complex workflows, and detailed reasoning sequences in a single cycle. Moreover, it is engineered for high-throughput, real-time performance, making it particularly adept at handling multi-turn dialogues, invoking tools, and executing agent-based workflows that involve intricate planning and reasoning tasks. This versatility positions Nemotron 3 Nano as a leading choice for applications requiring advanced cognitive capabilities.

MiniMax M2.7 is a powerful AI model built to drive real-world productivity across coding, search, and office-based workflows. It is trained using reinforcement learning across a wide range of real-world environments, enabling it to execute complex, multi-step tasks with precision and efficiency. The model demonstrates strong problem-solving capabilities by breaking down challenges into structured steps before generating solutions across multiple programming languages. It delivers high-speed performance with rapid token output, ensuring faster completion of demanding tasks. With optimized reasoning, it reduces token usage and execution time, making it more efficient than previous models. M2.7 also achieves state-of-the-art results in software engineering benchmarks, significantly improving response times for technical issues. Its advanced agentic capabilities allow it to work seamlessly with tools and support complex workflows with high skill accuracy. The model is designed to handle professional tasks, including multi-turn interactions and high-quality document editing. It also provides strong support for office productivity, enabling efficient handling of structured data and business tasks. With competitive pricing, it delivers high performance while remaining cost-effective. Overall, it combines speed, intelligence, and versatility to meet the needs of modern professionals and teams.

GLM-4.5V is an evolution of the GLM-4.5-Air model, incorporating a Mixture-of-Experts (MoE) framework that boasts a remarkable total of 106 billion parameters, with 12 billion specifically dedicated to activation. This model stands out by delivering top-tier performance among open-source vision-language models (VLMs) of comparable scale, demonstrating exceptional capabilities across 42 public benchmarks in diverse contexts such as images, videos, documents, and GUI interactions. It offers an extensive array of multimodal functionalities, encompassing image reasoning tasks like scene understanding, spatial recognition, and multi-image analysis, alongside video comprehension tasks that include segmentation and event recognition. Furthermore, it excels in parsing complex charts and lengthy documents, facilitating GUI-agent workflows through tasks like screen reading and desktop automation, while also providing accurate visual grounding by locating objects and generating bounding boxes. Additionally, the introduction of a "Thinking Mode" switch enhances user experience by allowing the selection of either rapid responses or more thoughtful reasoning based on the situation at hand. This innovative feature makes GLM-4.5V not only versatile but also adaptable to various user needs.