Best Foundation Models for Hugging Face
View:
Compare the Top Foundation Models that integrate with Hugging Face as of August 2026
Sort By:
This a list of Foundation Models that integrate with Hugging Face. Use the filters on the left to add additional filters for products that have integrations with Hugging Face. View the products that work with Hugging Face in the table below.
What are Foundation Models for Hugging Face?
Foundation models are large-scale artificial intelligence models trained on vast and diverse datasets that serve as the underlying technology for a wide range of AI applications. These models learn general-purpose capabilities such as language understanding, reasoning, image recognition, code generation, speech processing, and multimodal comprehension, allowing them to be adapted or fine-tuned for specific tasks across industries. Foundation models power applications including chatbots, AI agents, search, content generation, software development, scientific research, and business automation. Many are available through cloud APIs, open-source distributions, and enterprise AI platforms, supporting custom model development, retrieval-augmented generation (RAG), and domain-specific optimization. By providing reusable, general-purpose intelligence, foundation models enable organizations to accelerate AI development, reduce implementation costs, and build sophisticated AI-powered applications. Compare and read user reviews of the best Foundation Models for Hugging Face currently available using the table below. This list is updated regularly.
-
1
Qwen3.8-Max
Alibaba
Qwen3.8-Max is Qwen’s most capable model to date, built as a Max-class AI model for coding, work, research, long-horizon tasks, and multimodal agents. It scales to 2.4 trillion parameters with 95 billion active parameters and is available through QwenCloud. The model is designed to complete complex, open-ended tasks end to end with greater reliability and minimal human involvement. Qwen3.8-Max supports autonomous coding workflows, agentic development, research reproduction, visual reasoning, document understanding, video analysis, and real-world productivity tasks. It can integrate with popular agent frameworks and coding assistants, including Claude Code, Codex, Qoder CLI, Qwen Code, and OpenClaw. Built for developers, researchers, enterprises, and AI agent builders, Qwen3.8-Max helps teams automate sophisticated work across code, documents, tools, interfaces, and multimodal content.Starting Price: $2 per 1M (input) -
2
Qwen
Alibaba
Qwen is a powerful, free AI assistant built on the advanced Qwen model series, designed to help anyone with creativity, research, problem-solving, and everyday tasks. While Qwen Chat is the main interface for most users, Qwen itself powers a broad range of intelligent capabilities including image generation, deep research, website creation, advanced reasoning, and context-aware search. Its multimodal intelligence enables Qwen to understand and process text, images, audio, and video simultaneously for richer insights. Qwen is available on web, desktop, and mobile, ensuring seamless access across all devices. For developers, the Qwen API provides OpenAI-compatible endpoints, making integration simple and allowing Qwen’s intelligence to power apps, services, and automation. Whether you're chatting through Qwen Chat or building with the Qwen API, Qwen delivers fast, flexible, and highly capable AI support.Starting Price: Free -
3
Qwen3.8-27B
Alibaba
Qwen3.8-27B is a compact open-weights model in Alibaba’s Qwen3.8 family, aimed at developers and researchers who want strong local AI performance without using the full Max-scale model. Reports from Alibaba’s Qwen3.8 launch state that Qwen3.8-27B was planned for open-weight release alongside Qwen3.8-Max, expanding access for builders working on AI applications. The model is positioned for coding, research, professional workflows, and local deployment scenarios where a 27B model can be more practical than frontier-scale systems. Qwen3.8’s broader launch emphasizes software development, document processing, data analysis, and professional “cowork” use cases. Qwen3.8-27B is especially relevant for teams that need a capable open model for experimentation, coding agents, assistant workflows, and self-hosted inference. Built for practical deployment, Qwen3.8-27B gives developers a smaller Qwen3.8 option for building AI tools, testing agents, and running advanced language model workflows. -
4
Qwen-7B
Alibaba
Qwen-7B is the 7B-parameter version of the large language model series, Qwen (abbr. Tongyi Qianwen), proposed by Alibaba Cloud. Qwen-7B is a Transformer-based large language model, which is pretrained on a large volume of data, including web texts, books, codes, etc. Additionally, based on the pretrained Qwen-7B, we release Qwen-7B-Chat, a large-model-based AI assistant, which is trained with alignment techniques. The features of the Qwen-7B series include: Trained with high-quality pretraining data. We have pretrained Qwen-7B on a self-constructed large-scale high-quality dataset of over 2.2 trillion tokens. The dataset includes plain texts and codes, and it covers a wide range of domains, including general domain data and professional domain data. Strong performance. In comparison with the models of the similar model size, we outperform the competitors on a series of benchmark datasets, which evaluates natural language understanding, mathematics, coding, etc. And more.Starting Price: Free -
5
Codestral Mamba
Mistral AI
As a tribute to Cleopatra, whose glorious destiny ended in tragic snake circumstances, we are proud to release Codestral Mamba, a Mamba2 language model specialized in code generation, available under an Apache 2.0 license. Codestral Mamba is another step in our effort to study and provide new architectures. It is available for free use, modification, and distribution, and we hope it will open new perspectives in architecture research. Mamba models offer the advantage of linear time inference and the theoretical ability to model sequences of infinite length. It allows users to engage with the model extensively with quick responses, irrespective of the input length. This efficiency is especially relevant for code productivity use cases, this is why we trained this model with advanced code and reasoning capabilities, enabling it to perform on par with SOTA transformer-based models.Starting Price: Free -
6
Mathstral
Mistral AI
As a tribute to Archimedes, whose 2311th anniversary we’re celebrating this year, we are proud to release our first Mathstral model, a specific 7B model designed for math reasoning and scientific discovery. The model has a 32k context window published under the Apache 2.0 license. We’re contributing Mathstral to the science community to bolster efforts in advanced mathematical problems requiring complex, multi-step logical reasoning. The Mathstral release is part of our broader effort to support academic projects, it was produced in the context of our collaboration with Project Numina. Akin to Isaac Newton in his time, Mathstral stands on the shoulders of Mistral 7B and specializes in STEM subjects. It achieves state-of-the-art reasoning capacities in its size category across various industry-standard benchmarks. In particular, it achieves 56.6% on MATH and 63.47% on MMLU, with the following MMLU performance difference by subject between Mathstral 7B and Mistral 7B.Starting Price: Free -
7
Kimi K2 Thinking
Moonshot AI
Kimi K2 Thinking is an advanced open source reasoning model developed by Moonshot AI, designed specifically for long-horizon, multi-step workflows where the system interleaves chain-of-thought processes with tool invocation across hundreds of sequential tasks. The model uses a mixture-of-experts architecture with a total of 1 trillion parameters, yet only about 32 billion parameters are activated per inference pass, optimizing efficiency while maintaining vast capacity. It supports a context window of up to 256,000 tokens, enabling the handling of extremely long inputs and reasoning chains without losing coherence. Native INT4 quantization is built in, which reduces inference latency and memory usage without performance degradation. Kimi K2 Thinking is explicitly built for agentic workflows; it can autonomously call external tools, manage sequential logic steps (up to and typically between 200-300 tool calls in a single chain), and maintain consistent reasoning.Starting Price: Free -
8
CodeQwen
Alibaba
CodeQwen is the code version of Qwen, the large language model series developed by the Qwen team, Alibaba Cloud. It is a transformer-based decoder-only language model pre-trained on a large amount of data of codes. Strong code generation capabilities and competitive performance across a series of benchmarks. Supporting long context understanding and generation with the context length of 64K tokens. CodeQwen supports 92 coding languages and provides excellent performance in text-to-SQL, bug fixes, etc. You can just write several lines of code with transformers to chat with CodeQwen. Essentially, we build the tokenizer and the model from pre-trained methods, and we use the generate method to perform chatting with the help of the chat template provided by the tokenizer. We apply the ChatML template for chat models following our previous practice. The model completes the code snippets according to the given prompts, without any additional formatting.Starting Price: Free -
9
Qwen2
Alibaba
Qwen2 is the large language model series developed by Qwen team, Alibaba Cloud. Qwen2 is a series of large language models developed by the Qwen team at Alibaba Cloud. It includes both base language models and instruction-tuned models, ranging from 0.5 billion to 72 billion parameters, and features both dense models and a Mixture-of-Experts model. The Qwen2 series is designed to surpass most previous open-weight models, including its predecessor Qwen1.5, and to compete with proprietary models across a broad spectrum of benchmarks in language understanding, generation, multilingual capabilities, coding, mathematics, and reasoning.Starting Price: Free -
10
Qwen2.5-Max
Alibaba
Qwen2.5-Max is a large-scale Mixture-of-Experts (MoE) model developed by the Qwen team, pretrained on over 20 trillion tokens and further refined through Supervised Fine-Tuning (SFT) and Reinforcement Learning from Human Feedback (RLHF). In evaluations, it outperforms models like DeepSeek V3 in benchmarks such as Arena-Hard, LiveBench, LiveCodeBench, and GPQA-Diamond, while also demonstrating competitive results in other assessments, including MMLU-Pro. Qwen2.5-Max is accessible via API through Alibaba Cloud and can be explored interactively on Qwen Chat.Starting Price: Free -
11
Qwen2.5-VL
Alibaba
Qwen2.5-VL is the latest vision-language model from the Qwen series, representing a significant advancement over its predecessor, Qwen2-VL. This model excels in visual understanding, capable of recognizing a wide array of objects, including text, charts, icons, graphics, and layouts within images. It functions as a visual agent, capable of reasoning and dynamically directing tools, enabling applications such as computer and phone usage. Qwen2.5-VL can comprehend videos exceeding one hour in length and can pinpoint relevant segments within them. Additionally, it accurately localizes objects in images by generating bounding boxes or points and provides stable JSON outputs for coordinates and attributes. The model also supports structured outputs for data like scanned invoices, forms, and tables, benefiting sectors such as finance and commerce. Available in base and instruct versions across 3B, 7B, and 72B sizes, Qwen2.5-VL is accessible through platforms like Hugging Face and ModelScope.Starting Price: Free -
12
NVIDIA Cosmos
NVIDIA
NVIDIA Cosmos is a developer-first platform of state-of-the-art generative World Foundation Models (WFMs), advanced video tokenizers, guardrails, and an accelerated data processing and curation pipeline designed to supercharge physical AI development. It enables developers working on autonomous vehicles, robotics, and video analytics AI agents to generate photorealistic, physics-aware synthetic video data, trained on an immense dataset including 20 million hours of real-world and simulated video, to rapidly simulate future scenarios, train world models, and fine‑tune custom behaviors. It includes three core WFM types; Cosmos Predict, capable of generating up to 30 seconds of continuous video from multimodal inputs; Cosmos Transfer, which adapts simulations across environments and lighting for versatile domain augmentation; and Cosmos Reason, a vision-language model that applies structured reasoning to interpret spatial-temporal data for planning and decision-making.Starting Price: Free -
13
DeepSeek-V3.2-Exp
DeepSeek
Introducing DeepSeek-V3.2-Exp, our latest experimental model built on V3.1-Terminus, debuting DeepSeek Sparse Attention (DSA) for faster and more efficient inference and training on long contexts. DSA enables fine-grained sparse attention with minimal loss in output quality, boosting performance for long-context tasks while reducing compute costs. Benchmarks indicate that V3.2-Exp performs on par with V3.1-Terminus despite these efficiency gains. The model is now live across app, web, and API. Alongside this, the DeepSeek API prices have been cut by over 50% immediately to make access more affordable. For a transitional period, users can still access V3.1-Terminus via a temporary API endpoint until October 15, 2025. DeepSeek welcomes feedback on DSA via its feedback portal. In conjunction with the release, DeepSeek-V3.2-Exp has been open-sourced: the model weights and supporting technology (including key GPU kernels in TileLang and CUDA) are available on Hugging Face.Starting Price: Free -
14
DeepSeek-V3.2
DeepSeek
DeepSeek-V3.2 is a next-generation open large language model designed for efficient reasoning, complex problem solving, and advanced agentic behavior. It introduces DeepSeek Sparse Attention (DSA), a long-context attention mechanism that dramatically reduces computation while preserving performance. The model is trained with a scalable reinforcement learning framework, allowing it to achieve results competitive with GPT-5 and even surpass it in its Speciale variant. DeepSeek-V3.2 also includes a large-scale agent task synthesis pipeline that generates structured reasoning and tool-use demonstrations for post-training. The model features an updated chat template with new tool-calling logic and the optional developer role for agent workflows. With gold-medal performance in the IMO and IOI 2025 competitions, DeepSeek-V3.2 demonstrates elite reasoning capabilities for both research and applied AI scenarios.Starting Price: Free -
15
DeepSeek-V3.2-Speciale
DeepSeek
DeepSeek-V3.2-Speciale is a high-compute variant of the DeepSeek-V3.2 model, created specifically for deep reasoning and advanced problem-solving tasks. It builds on DeepSeek Sparse Attention (DSA), a custom long-context attention mechanism that reduces computational overhead while preserving high performance. Through a large-scale reinforcement learning framework and extensive post-training compute, the Speciale variant surpasses GPT-5 on reasoning benchmarks and matches the capabilities of Gemini-3.0-Pro. The model achieved gold-medal performance in the International Mathematical Olympiad (IMO) 2025 and International Olympiad in Informatics (IOI) 2025. DeepSeek-V3.2-Speciale does not support tool-calling, making it purely optimized for uninterrupted reasoning and analytical accuracy. Released under the MIT license, it provides researchers and developers an open, state-of-the-art model focused entirely on high-precision reasoning.Starting Price: Free -
16
MiMo-V2-Flash
Xiaomi Technology
MiMo-V2-Flash is an open weight large language model developed by Xiaomi based on a Mixture-of-Experts (MoE) architecture that blends high performance with inference efficiency. It has 309 billion total parameters but activates only 15 billion active parameters per inference, letting it balance reasoning quality and computational efficiency while supporting extremely long context handling, for tasks like long-document understanding, code generation, and multi-step agent workflows. It incorporates a hybrid attention mechanism that interleaves sliding-window and global attention layers to reduce memory usage and maintain long-range comprehension, and it uses a Multi-Token Prediction (MTP) design that accelerates inference by processing batches of tokens in parallel. MiMo-V2-Flash delivers very fast generation speeds (up to ~150 tokens/second) and is optimized for agentic applications requiring sustained reasoning and multi-turn interactions.Starting Price: Free -
17
Step 3.5 Flash
StepFun
Step 3.5 Flash is an advanced open source foundation language model engineered for frontier reasoning and agentic capabilities with exceptional efficiency, built on a sparse Mixture of Experts (MoE) architecture that selectively activates only about 11 billion of its ~196 billion parameters per token to deliver high-density intelligence and real-time responsiveness. Its 3-way Multi-Token Prediction (MTP-3) enables generation throughput in the hundreds of tokens per second for complex multi-step reasoning chains and task execution, and it supports efficient long contexts with a hybrid sliding window attention approach that reduces computational overhead across large datasets or codebases. It demonstrates robust performance on benchmarks for reasoning, coding, and agentic tasks, rivaling or exceeding many larger proprietary models, and includes a scalable reinforcement learning framework for consistent self-improvement.Starting Price: Free -
18
Qwen3.6-35B-A3B
Alibaba
Qwen3.5-35B-A3B is part of the Qwen3.5 “Medium” model series, designed as a highly efficient, multimodal foundation model that balances strong reasoning ability with practical deployment requirements. It uses a Mixture-of-Experts (MoE) architecture with 35 billion total parameters but activates only about 3 billion per token, allowing it to deliver performance comparable to much larger models while significantly reducing computational cost. The model integrates a hybrid attention mechanism that combines linear attention with standard attention layers, enabling efficient long-context processing and improved scalability for complex tasks. As a native vision-language model, it can process both text and visual inputs, supporting use cases such as multimodal reasoning, coding, and agent-based workflows. It is designed to function as a general-purpose “AI agent,” capable of planning, tool use, and structured problem solving rather than just conversational responses.Starting Price: Free -
19
MiniMax M1
MiniMax
MiniMax‑M1 is a large‑scale hybrid‑attention reasoning model released by MiniMax AI under the Apache 2.0 license. It supports an unprecedented 1 million‑token context window and up to 80,000-token outputs, enabling extended reasoning across long documents. Trained using large‑scale reinforcement learning with a novel CISPO algorithm, MiniMax‑M1 completed full training on 512 H800 GPUs in about three weeks. It achieves state‑of‑the‑art performance on benchmarks in mathematics, coding, software engineering, tool usage, and long‑context understanding, matching or outperforming leading models. Two model variants are available (40K and 80K thinking budgets), with weights and deployment scripts provided via GitHub and Hugging Face. -
20
Qwen3.6
Alibaba
Qwen3.6 is a large language model developed by Alibaba as part of its Qwen AI model family, designed for real-world applications and advanced reasoning tasks. It focuses on improving stability, usability, and performance compared to earlier versions. The model supports multimodal capabilities, allowing it to process and reason across text, images, and other data types. Qwen3.6 is particularly strong in coding and developer workflows, offering improved accuracy for complex programming tasks. It uses a mixture-of-experts architecture, enabling efficient performance while maintaining large-scale model capabilities. The model is designed to be deployable in production environments, including enterprise and cloud-based systems. It can be integrated into applications or run locally using open-weight variants. Overall, Qwen3.6 delivers a powerful, efficient, and versatile AI solution for modern use cases.Starting Price: Free -
21
Evo 2
Arc Institute
Evo 2 is a genomic foundation model capable of generalist prediction and design tasks across DNA, RNA, and proteins. It utilizes a frontier deep learning architecture to model biological sequences at single-nucleotide resolution, achieving near-linear scaling of compute and memory relative to context length. Trained with 40 billion parameters and a 1 megabase context length, Evo 2 processes over 9 trillion nucleotides from diverse eukaryotic and prokaryotic genomes. This extensive training enables Evo 2 to perform zero-shot function prediction across multiple biological modalities, including DNA, RNA, and proteins, and to generate novel sequences with plausible genomic architecture. The model's capabilities have been demonstrated in tasks such as designing functional CRISPR systems and predicting disease-causing mutations in human genes. Evo 2 is publicly accessible via Arc's GitHub repository and is integrated into the NVIDIA BioNeMo framework. -
22
Qwen3.8-2.4T-A95B
Alibaba
Qwen3.8-2.4T-A95B is the largest open model in the Qwen3.8 family, bringing Qwen-Max-class capabilities to an open release. Built on the architectural foundation of Qwen3.5, it delivers substantial improvements across coding, professional work, research, and long-horizon agentic tasks, with a focus on carrying complex, multi-step work through to completion more reliably. The causal language model uses a mixture-of-experts architecture with 2.4 trillion total parameters and 95 billion activated parameters, including 512 experts with 10 routed and one shared expert active at a time. It supports a native context length of 262,144 tokens that can be extended to approximately 1.01 million tokens. Agent execution is strengthened through better autonomous planning and improved handling of environment feedback, while broader compatibility with popular agent harnesses and development tools simplifies integration into existing stacks.
- Previous
- You're on page 1
- Next