Alternatives to Tülu 3

Molmo represents a cutting-edge family of multimodal AI models crafted by the Allen Institute for AI (Ai2). These innovative models are specifically engineered to connect the divide between open-source and proprietary systems, ensuring they perform competitively across numerous academic benchmarks and assessments by humans. In contrast to many existing multimodal systems that depend on synthetic data sourced from proprietary frameworks, Molmo is exclusively trained on openly available data, which promotes transparency and reproducibility in AI research. A significant breakthrough in the development of Molmo is the incorporation of PixMo, a unique dataset filled with intricately detailed image captions gathered from human annotators who utilized speech-based descriptions, along with 2D pointing data that empowers the models to respond to inquiries with both natural language and non-verbal signals. This capability allows Molmo to engage with its surroundings in a more sophisticated manner, such as by pointing to specific objects within images, thereby broadening its potential applications in diverse fields, including robotics, augmented reality, and interactive user interfaces. Furthermore, the advancements made by Molmo set a new standard for future multimodal AI research and application 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.

Code Llama is an advanced language model designed to generate code through text prompts, distinguishing itself as a leading tool among publicly accessible models for coding tasks. This innovative model not only streamlines workflows for existing developers but also aids beginners in overcoming challenges associated with learning to code. Its versatility positions Code Llama as both a valuable productivity enhancer and an educational resource, assisting programmers in creating more robust and well-documented software solutions. Additionally, users can generate both code and natural language explanations by providing either type of prompt, making it an adaptable tool for various programming needs. Available for free for both research and commercial applications, Code Llama is built upon Llama 2 architecture and comes in three distinct versions: the foundational Code Llama model, Code Llama - Python which is tailored specifically for Python programming, and Code Llama - Instruct, optimized for comprehending and executing natural language directives effectively.

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.

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.

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.

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

OpenPipe offers an efficient platform for developers to fine-tune their models. It allows you to keep your datasets, models, and evaluations organized in a single location. You can train new models effortlessly with just a click. The system automatically logs all LLM requests and responses for easy reference. You can create datasets from the data you've captured, and even train multiple base models using the same dataset simultaneously. Our managed endpoints are designed to handle millions of requests seamlessly. Additionally, you can write evaluations and compare the outputs of different models side by side for better insights. A few simple lines of code can get you started; just swap out your Python or Javascript OpenAI SDK with an OpenPipe API key. Enhance the searchability of your data by using custom tags. Notably, smaller specialized models are significantly cheaper to operate compared to large multipurpose LLMs. Transitioning from prompts to models can be achieved in minutes instead of weeks. Our fine-tuned Mistral and Llama 2 models routinely exceed the performance of GPT-4-1106-Turbo, while also being more cost-effective. With a commitment to open-source, we provide access to many of the base models we utilize. When you fine-tune Mistral and Llama 2, you maintain ownership of your weights and can download them whenever needed. Embrace the future of model training and deployment with OpenPipe's comprehensive tools and features.

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.

Tinker is an innovative training API tailored for researchers and developers, providing comprehensive control over model fine-tuning while simplifying the complexities of infrastructure management. It offers essential primitives that empower users to create bespoke training loops, supervision techniques, and reinforcement learning workflows. Currently, it facilitates LoRA fine-tuning on open-weight models from both the LLama and Qwen families, accommodating a range of model sizes from smaller variants to extensive mixture-of-experts configurations. Users can write Python scripts to manage data, loss functions, and algorithmic processes, while Tinker autonomously takes care of scheduling, resource distribution, distributed training, and recovery from failures. The platform allows users to download model weights at various checkpoints without the burden of managing the computational environment. Delivered as a managed service, Tinker executes training jobs on Thinking Machines’ proprietary GPU infrastructure, alleviating users from the challenges of cluster orchestration and enabling them to focus on building and optimizing their models. This seamless integration of capabilities makes Tinker a vital tool for advancing machine learning research and development.

Phi-4-reasoning is an advanced transformer model featuring 14 billion parameters, specifically tailored for tackling intricate reasoning challenges, including mathematics, programming, algorithm development, and strategic planning. Through a meticulous process of supervised fine-tuning on select "teachable" prompts and reasoning examples created using o3-mini, it excels at generating thorough reasoning sequences that optimize computational resources during inference. By integrating outcome-driven reinforcement learning, Phi-4-reasoning is capable of producing extended reasoning paths. Its performance notably surpasses that of significantly larger open-weight models like DeepSeek-R1-Distill-Llama-70B and nears the capabilities of the comprehensive DeepSeek-R1 model across various reasoning applications. Designed for use in settings with limited computing power or high latency, Phi-4-reasoning is fine-tuned with synthetic data provided by DeepSeek-R1, ensuring it delivers precise and methodical problem-solving. This model's ability to handle complex tasks with efficiency makes it a valuable tool in numerous computational contexts.

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.

StableVicuna represents the inaugural large-scale open-source chatbot developed through reinforced learning from human feedback (RLHF). It is an advanced version of the Vicuna v0 13b model, which has undergone further instruction fine-tuning and RLHF training. To attain the impressive capabilities of StableVicuna, we use Vicuna as the foundational model and adhere to the established three-stage RLHF framework proposed by Steinnon et al. and Ouyang et al. Specifically, we perform additional training on the base Vicuna model with supervised fine-tuning (SFT), utilizing a blend of three distinct datasets. The first is the OpenAssistant Conversations Dataset (OASST1), which consists of 161,443 human-generated messages across 66,497 conversation trees in 35 languages. The second dataset is GPT4All Prompt Generations, encompassing 437,605 prompts paired with responses created by GPT-3.5 Turbo. Lastly, the Alpaca dataset features 52,000 instructions and demonstrations that were produced using OpenAI's text-davinci-003 model. This collective approach to training enhances the chatbot's ability to engage effectively in diverse conversational contexts.

Instruction-following models like GPT-3.5 (text-DaVinci-003), ChatGPT, Claude, and Bing Chat have seen significant advancements in their capabilities, leading to a rise in their usage among individuals in both personal and professional contexts. Despite their growing popularity and integration into daily tasks, these models are not without their shortcomings, as they can sometimes disseminate inaccurate information, reinforce harmful stereotypes, and use inappropriate language. To effectively tackle these critical issues, it is essential for researchers and scholars to become actively involved in exploring these models further. However, conducting research on instruction-following models within academic settings has posed challenges due to the unavailability of models with comparable functionality to proprietary options like OpenAI’s text-DaVinci-003. In response to this gap, we are presenting our insights on an instruction-following language model named Alpaca, which has been fine-tuned from Meta’s LLaMA 7B model, aiming to contribute to the discourse and development in this field. This initiative represents a step towards enhancing the understanding and capabilities of instruction-following models in a more accessible manner for researchers.

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.

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.

EXAONE Deep represents a collection of advanced language models that are enhanced for reasoning, created by LG AI Research, and come in sizes of 2.4 billion, 7.8 billion, and 32 billion parameters. These models excel in a variety of reasoning challenges, particularly in areas such as mathematics and coding assessments. Significantly, the EXAONE Deep 2.4B model outshines other models of its size, while the 7.8B variant outperforms both open-weight models of similar dimensions and the proprietary reasoning model known as OpenAI o1-mini. Furthermore, the EXAONE Deep 32B model competes effectively with top-tier open-weight models in the field. The accompanying repository offers extensive documentation that includes performance assessments, quick-start guides for leveraging EXAONE Deep models with the Transformers library, detailed explanations of quantized EXAONE Deep weights formatted in AWQ and GGUF, as well as guidance on how to run these models locally through platforms like llama.cpp and Ollama. Additionally, this resource serves to enhance user understanding and accessibility to the capabilities of EXAONE Deep models.

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

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.

Vicuna-13B is an open-source conversational agent developed through the fine-tuning of LLaMA, utilizing a dataset of user-shared dialogues gathered from ShareGPT. Initial assessments, with GPT-4 serving as an evaluator, indicate that Vicuna-13B achieves over 90% of the quality exhibited by OpenAI's ChatGPT and Google Bard, and it surpasses other models such as LLaMA and Stanford Alpaca in more than 90% of instances. The entire training process for Vicuna-13B incurs an estimated expenditure of approximately $300. Additionally, the source code and model weights, along with an interactive demonstration, are made available for public access under non-commercial terms, fostering a collaborative environment for further development and exploration. This openness encourages innovation and enables users to experiment with the model's capabilities in diverse applications.

Open Coding Agents represent a suite of fully open, high-performance AI coding models along with a training methodology introduced by the Allen Institute for AI, designed to simplify the process of creating, customizing, and training coding agents across various repositories in an accessible, cost-effective, and transparent manner; this platform encompasses models, code, training recipes, and tools that can be activated with minimal configuration, allowing users to adapt agents to their specific codebases and engineering practices for a variety of tasks including code generation, code review, debugging, maintenance, and code explanation. By departing from conventional closed and costly systems, these agents provide an open pipeline that extends from models to training data, facilitating fine-tuning on internal code, which helps agents learn about organization-specific APIs, patterns, and workflows; the inaugural release, SERA (Soft-verified Efficient Repository Agents), sets a new standard in coding benchmarks while maintaining a significantly lower compute cost than typical solutions, showcasing the potential for innovation in the field of AI-driven coding. As the landscape of coding becomes increasingly complex, the introduction of such models promises to democratize access to advanced coding assistance, paving the way for a more efficient development process.

Push the limits of individual alignment, artificial consciousness, open-source software, and decentralization through experimentation that larger corporations and governments often shy away from. Hermes 3 features sophisticated long-term context retention, the ability to engage in multi-turn conversations, and intricate roleplaying and internal monologue capabilities, alongside improved functionality for agentic function-calling. The design of this model emphasizes precise adherence to system prompts and instruction sets in a flexible way. By fine-tuning Llama 3.1 across various scales, including 8B, 70B, and 405B, and utilizing a dataset largely composed of synthetically generated inputs, Hermes 3 showcases performance that rivals and even surpasses Llama 3.1, while also unlocking greater potential in reasoning and creative tasks. This series of instructive and tool-utilizing models exhibits exceptional reasoning and imaginative skills, paving the way for innovative applications. Ultimately, Hermes 3 represents a significant advancement in the landscape of AI development.

Oumi is an entirely open-source platform that enhances the complete lifecycle of foundation models, encompassing everything from data preparation and training to evaluation and deployment. It facilitates the training and fine-tuning of models with parameter counts ranging from 10 million to an impressive 405 billion, utilizing cutting-edge methodologies such as SFT, LoRA, QLoRA, and DPO. Supporting both text-based and multimodal models, Oumi is compatible with various architectures like Llama, DeepSeek, Qwen, and Phi. The platform also includes tools for data synthesis and curation, allowing users to efficiently create and manage their training datasets. For deployment, Oumi seamlessly integrates with well-known inference engines such as vLLM and SGLang, which optimizes model serving. Additionally, it features thorough evaluation tools across standard benchmarks to accurately measure model performance. Oumi's design prioritizes flexibility, enabling it to operate in diverse environments ranging from personal laptops to powerful cloud solutions like AWS, Azure, GCP, and Lambda, making it a versatile choice for developers. This adaptability ensures that users can leverage the platform regardless of their operational context, enhancing its appeal across different use cases.

Liquid AI's LFM2.5 represents an advanced iteration of on-device AI foundation models, engineered to provide high-efficiency and performance for AI inference on edge devices like smartphones, laptops, vehicles, IoT systems, and embedded hardware without the need for cloud computing resources. This new version builds upon the earlier LFM2 framework by greatly enhancing the scale of pretraining and the stages of reinforcement learning, resulting in a suite of hybrid models that boast around 1.2 billion parameters while effectively balancing instruction adherence, reasoning skills, and multimodal functionalities for practical applications. The LFM2.5 series comprises various models including Base (for fine-tuning and personalization), Instruct (designed for general-purpose instruction), Japanese-optimized, Vision-Language, and Audio-Language variants, all meticulously crafted for rapid on-device inference even with stringent memory limitations. These models are also made available as open-weight options, facilitating deployment through platforms such as llama.cpp, MLX, vLLM, and ONNX, thus ensuring versatility for developers. With these enhancements, LFM2.5 positions itself as a robust solution for diverse AI-driven tasks in real-world environments.

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.

Stability AI, along with its CarperAI lab, is excited to unveil Stable Beluga 1 and its advanced successor, Stable Beluga 2, previously known as FreeWilly, both of which are robust new Large Language Models (LLMs) available for public use. These models exhibit remarkable reasoning capabilities across a wide range of benchmarks, showcasing their versatility and strength. Stable Beluga 1 is built on the original LLaMA 65B foundation model and has undergone meticulous fine-tuning with a novel synthetically-generated dataset utilizing Supervised Fine-Tune (SFT) in the conventional Alpaca format. In a similar vein, Stable Beluga 2 utilizes the LLaMA 2 70B foundation model, pushing the boundaries of performance in the industry. Their development marks a significant step forward in the evolution of open access AI technologies.

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.

Phi-4-mini-reasoning is a transformer-based language model with 3.8 billion parameters, specifically designed to excel in mathematical reasoning and methodical problem-solving within environments that have limited computational capacity or latency constraints. Its optimization stems from fine-tuning with synthetic data produced by the DeepSeek-R1 model, striking a balance between efficiency and sophisticated reasoning capabilities. With training that encompasses over one million varied math problems, ranging in complexity from middle school to Ph.D. level, Phi-4-mini-reasoning demonstrates superior performance to its base model in generating lengthy sentences across multiple assessments and outshines larger counterparts such as OpenThinker-7B, Llama-3.2-3B-instruct, and DeepSeek-R1. Equipped with a 128K-token context window, it also facilitates function calling, which allows for seamless integration with various external tools and APIs. Moreover, Phi-4-mini-reasoning can be quantized through the Microsoft Olive or Apple MLX Framework, enabling its deployment on a variety of edge devices, including IoT gadgets, laptops, and smartphones. Its design not only enhances user accessibility but also expands the potential for innovative applications in mathematical fields.

Nebius Token Factory is an advanced AI inference platform that enables the production of both open-source and proprietary AI models without the need for manual infrastructure oversight. It provides enterprise-level inference endpoints that ensure consistent performance, automatic scaling of throughput, and quick response times, even when faced with high request traffic. With a remarkable 99.9% uptime, it accommodates both unlimited and customized traffic patterns according to specific workload requirements, facilitating a seamless shift from testing to worldwide implementation. Supporting a diverse array of open-source models, including Llama, Qwen, DeepSeek, GPT-OSS, Flux, and many more, Nebius Token Factory allows teams to host and refine models via an intuitive API or dashboard interface. Users have the flexibility to upload LoRA adapters or fully fine-tuned versions directly, while still benefiting from the same enterprise-grade performance assurances for their custom models. This level of support ensures that organizations can confidently leverage AI technology to meet their evolving needs.

Kluster.ai is an AI cloud platform tailored for developers, enabling quick deployment, scaling, and fine-tuning of large language models (LLMs) with remarkable efficiency. Crafted by developers with a focus on developer needs, it features Adaptive Inference, a versatile service that dynamically adjusts to varying workload demands, guaranteeing optimal processing performance and reliable turnaround times. This Adaptive Inference service includes three unique processing modes: real-time inference for tasks requiring minimal latency, asynchronous inference for budget-friendly management of tasks with flexible timing, and batch inference for the streamlined processing of large volumes of data. It accommodates an array of innovative multimodal models for various applications such as chat, vision, and coding, featuring models like Meta's Llama 4 Maverick and Scout, Qwen3-235B-A22B, DeepSeek-R1, and Gemma 3. Additionally, Kluster.ai provides an OpenAI-compatible API, simplifying the integration of these advanced models into developers' applications, and thereby enhancing their overall capabilities. This platform ultimately empowers developers to harness the full potential of AI technologies in their projects.

Yi-Lightning, a product of 01.AI and spearheaded by Kai-Fu Lee, marks a significant leap forward in the realm of large language models, emphasizing both performance excellence and cost-effectiveness. With the ability to process a context length of up to 16K tokens, it offers an attractive pricing model of $0.14 per million tokens for both inputs and outputs, making it highly competitive in the market. The model employs an improved Mixture-of-Experts (MoE) framework, featuring detailed expert segmentation and sophisticated routing techniques that enhance its training and inference efficiency. Yi-Lightning has distinguished itself across multiple fields, achieving top distinctions in areas such as Chinese language processing, mathematics, coding tasks, and challenging prompts on chatbot platforms, where it ranked 6th overall and 9th in style control. Its creation involved an extensive combination of pre-training, targeted fine-tuning, and reinforcement learning derived from human feedback, which not only enhances its performance but also prioritizes user safety. Furthermore, the model's design includes significant advancements in optimizing both memory consumption and inference speed, positioning it as a formidable contender in its field.

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

Entry Point AI serves as a cutting-edge platform for optimizing both proprietary and open-source language models. It allows users to manage prompts, fine-tune models, and evaluate their performance all from a single interface. Once you hit the ceiling of what prompt engineering can achieve, transitioning to model fine-tuning becomes essential, and our platform simplifies this process. Rather than instructing a model on how to act, fine-tuning teaches it desired behaviors. This process works in tandem with prompt engineering and retrieval-augmented generation (RAG), enabling users to fully harness the capabilities of AI models. Through fine-tuning, you can enhance the quality of your prompts significantly. Consider it an advanced version of few-shot learning where key examples are integrated directly into the model. For more straightforward tasks, you have the option to train a lighter model that can match or exceed the performance of a more complex one, leading to reduced latency and cost. Additionally, you can configure your model to avoid certain responses for safety reasons, which helps safeguard your brand and ensures proper formatting. By incorporating examples into your dataset, you can also address edge cases and guide the behavior of the model, ensuring it meets your specific requirements effectively. This comprehensive approach ensures that you not only optimize performance but also maintain control over the model's responses.

DeepCoder, an entirely open-source model for code reasoning and generation, has been developed through a partnership between Agentica Project and Together AI. Leveraging the foundation of DeepSeek-R1-Distilled-Qwen-14B, it has undergone fine-tuning via distributed reinforcement learning, achieving a notable accuracy of 60.6% on LiveCodeBench, which marks an 8% enhancement over its predecessor. This level of performance rivals that of proprietary models like o3-mini (2025-01-031 Low) and o1, all while operating with only 14 billion parameters. The training process spanned 2.5 weeks on 32 H100 GPUs, utilizing a carefully curated dataset of approximately 24,000 coding challenges sourced from validated platforms, including TACO-Verified, PrimeIntellect SYNTHETIC-1, and submissions to LiveCodeBench. Each problem mandated a legitimate solution along with a minimum of five unit tests to guarantee reliability during reinforcement learning training. Furthermore, to effectively manage long-range context, DeepCoder incorporates strategies such as iterative context lengthening and overlong filtering, ensuring it remains adept at handling complex coding tasks. This innovative approach allows DeepCoder to maintain high standards of accuracy and reliability in its code generation capabilities.

Unsloth is an innovative open-source platform specifically crafted to enhance and expedite the fine-tuning and training process of Large Language Models (LLMs). This platform empowers users to develop customized models, such as ChatGPT, in just a single day, a remarkable reduction from the usual training time of 30 days, achieving speeds that can be up to 30 times faster than Flash Attention 2 (FA2) while significantly utilizing 90% less memory. It supports advanced fine-tuning methods like LoRA and QLoRA, facilitating effective customization for models including Mistral, Gemma, and Llama across its various versions. The impressive efficiency of Unsloth arises from the meticulous derivation of computationally demanding mathematical processes and the hand-coding of GPU kernels, which leads to substantial performance enhancements without necessitating any hardware upgrades. On a single GPU, Unsloth provides a tenfold increase in processing speed and can achieve up to 32 times improvement on multi-GPU setups compared to FA2, with its functionality extending to a range of NVIDIA GPUs from Tesla T4 to H100, while also being portable to AMD and Intel graphics cards. This versatility ensures that a wide array of users can take full advantage of Unsloth's capabilities, making it a compelling choice for those looking to push the boundaries of model training efficiency.

Capture production data. Evaluate outputs together and fine-tune the performance of your LLM. A detailed log overview will help you understand what is happening in production. Work with domain experts, product managers and engineers to create reliable model outputs. Track AI metrics, such as speed, token usage, and quality scores. Copilot automates model evaluations and improvements for your use cases. Create, manage, or optimize prompts for precise and relevant interactions between AI models and users. Compare fine-tuned models and foundation models to improve prompt performance. Build a fine-tuning dataset with your team. Create custom fine-tuning data to optimize model performance.

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.

DeepEval offers an intuitive open-source framework designed for the assessment and testing of large language model systems, similar to what Pytest does but tailored specifically for evaluating LLM outputs. It leverages cutting-edge research to measure various performance metrics, including G-Eval, hallucinations, answer relevancy, and RAGAS, utilizing LLMs and a range of other NLP models that operate directly on your local machine. This tool is versatile enough to support applications developed through methods like RAG, fine-tuning, LangChain, or LlamaIndex. By using DeepEval, you can systematically explore the best hyperparameters to enhance your RAG workflow, mitigate prompt drift, or confidently shift from OpenAI services to self-hosting your Llama2 model. Additionally, the framework features capabilities for synthetic dataset creation using advanced evolutionary techniques and integrates smoothly with well-known frameworks, making it an essential asset for efficient benchmarking and optimization of LLM systems. Its comprehensive nature ensures that developers can maximize the potential of their LLM applications across various contexts.

Z-Image is a family of open-source image generation foundation models created by Alibaba's Tongyi-MAI team, utilizing a Scalable Single-Stream Diffusion Transformer architecture to produce both photorealistic and imaginative images from textual descriptions with only 6 billion parameters, which enhances its efficiency compared to many larger models while maintaining competitive quality and responsiveness to instructions. This model family comprises several variants, including Z-Image-Turbo, a distilled version designed for rapid inference that achieves results with as few as eight function evaluations and sub-second generation times on compatible GPUs; Z-Image, the comprehensive foundation model tailored for high-fidelity creative outputs and fine-tuning processes; Z-Image-Omni-Base, a flexible base checkpoint aimed at fostering community-driven advancements; and Z-Image-Edit, specifically optimized for image-to-image editing tasks while demonstrating strong adherence to instructions. Each variant of Z-Image serves distinct purposes, catering to a wide range of user needs within the realm of image generation.

Solar Mini is an advanced pre-trained large language model that matches the performance of GPT-3.5 while providing responses 2.5 times faster, all while maintaining a parameter count of under 30 billion. In December 2023, it secured the top position on the Hugging Face Open LLM Leaderboard by integrating a 32-layer Llama 2 framework, which was initialized with superior Mistral 7B weights, coupled with a novel method known as "depth up-scaling" (DUS) that enhances the model's depth efficiently without the need for intricate modules. Following the DUS implementation, the model undergoes further pretraining to restore and boost its performance, and it also includes instruction tuning in a question-and-answer format, particularly tailored for Korean, which sharpens its responsiveness to user prompts, while alignment tuning ensures its outputs align with human or sophisticated AI preferences. Solar Mini consistently surpasses rivals like Llama 2, Mistral 7B, Ko-Alpaca, and KULLM across a range of benchmarks, demonstrating that a smaller model can still deliver exceptional performance. This showcases the potential of innovative architectural strategies in the development of highly efficient AI models.

Dolly is an economical large language model that surprisingly demonstrates a notable level of instruction-following abilities similar to those seen in ChatGPT. While the Alpaca team's research revealed that cutting-edge models could be encouraged to excel in high-quality instruction adherence, our findings indicate that even older open-source models with earlier architectures can display remarkable behaviors when fine-tuned on a modest set of instructional training data. By utilizing an existing open-source model with 6 billion parameters from EleutherAI, Dolly has been slightly adjusted to enhance its ability to follow instructions, showcasing skills like brainstorming and generating text that were absent in its original form. This approach not only highlights the potential of older models but also opens new avenues for leveraging existing technologies in innovative ways.

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.

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.

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

Nexusflow has unveiled Athene-V2, its newest model suite boasting 72 billion parameters, which has been meticulously fine-tuned from Qwen 2.5 72B to rival the capabilities of GPT-4o. Within this suite, Athene-V2-Chat-72B stands out as a cutting-edge chat model that performs comparably to GPT-4o across various benchmarks; it excels particularly in chat helpfulness (Arena-Hard), ranks second in the code completion category on bigcode-bench-hard, and demonstrates strong abilities in mathematics (MATH) and accurate long log extraction. Furthermore, Athene-V2-Agent-72B seamlessly integrates chat and agent features, delivering clear and directive responses while surpassing GPT-4o in Nexus-V2 function calling benchmarks, specifically tailored for intricate enterprise-level scenarios. These innovations highlight a significant industry transition from merely increasing model sizes to focusing on specialized customization, showcasing how targeted post-training techniques can effectively enhance models for specific skills and applications. As technology continues to evolve, it becomes essential for developers to leverage these advancements to create increasingly sophisticated AI solutions.