Alternatives to TabFM
Compare TabFM alternatives for your business or organization using the curated list below. SourceForge ranks the best alternatives to TabFM in 2026. Compare features, ratings, user reviews, pricing, and more from TabFM competitors and alternatives in order to make an informed decision for your business.
-
1
Muse Spark Open
Meta
Muse Spark Open is an open-weight, natively multimodal language model for developers and organizations building AI applications, coding systems, and agentic workflows. The released model has 816 billion total parameters, 42 billion active parameters, and a 1-million-token context window. Teams can download the model weights, run inference through supported engines such as vLLM, SGLang, TensorRT-LLM, Hugging Face Transformers, Apple MLX, llama.cpp, or Unsloth, and customize the model using PyTorch. It supports long-context reasoning, software engineering, coding, tool use, and multimodal inputs. Deployment options include cloud inference through listed API providers or self-hosting on compatible GPU systems; Meta documents a desktop setup using two Apple M5 Ultra Mac Studio computers. The model is released under the Muse Model License with limited commercial terms.Starting Price: $0 -
2
Inkling
Thinking Machines Lab
Inkling is an open-weights multimodal AI model from Thinking Machines designed as a customizable foundation model for developers, researchers, and enterprises. The model is a Mixture-of-Experts transformer with 975 billion total parameters, 41 billion active parameters, and support for context windows up to 1 million tokens. Inkling was trained from scratch on text, images, audio, and video, giving it native capabilities across reasoning, coding, agentic tool use, vision, audio, factuality, and instruction following. It is built with controllable thinking effort so users can balance performance, latency, and token efficiency for different workloads. The model is available for fine-tuning on Tinker, with playground access, API availability through ecosystem partners, and full weights published on Hugging Face. Built for customization, Inkling gives teams an open-weights base model for building domain-specific AI systems, multimodal agents, coding workflows, research tools, and more.Starting Price: Free -
3
TabPFN-3.5
Prior Labs
TabPFN-3.5 is a tabular foundation model built for state-of-the-art predictions on structured data. It supports a wide range of prediction tasks, including churn, fraud, pricing, demand forecasting, risk, and other real-world data science problems, allowing teams to serve multiple use cases with one model. The model works with data as it is, handling missing values, outliers, categorical features, multi-table datasets, free text as a feature, thousands of distinct IDs without encoding, and hundreds of measurements per row. Users can feed in raw data, skip feature engineering and preprocessing, and get production-grade predictions from the first predict call. TabPFN-3.5 performs predictions in a single forward pass and is designed for both accuracy and speed, with fast inference for latency-critical predictive workflows. It supports production-scale datasets of up to one million rows natively and delivers 20x faster inference than previous model versions. -
4
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. -
5
Amazon SageMaker Autopilot
Amazon
Amazon SageMaker Autopilot eliminates the heavy lifting of building ML models. You simply provide a tabular dataset and select the target column to predict, and SageMaker Autopilot will automatically explore different solutions to find the best model. You then can directly deploy the model to production with just one click or iterate on the recommended solutions to further improve the model quality. You can use Amazon SageMaker Autopilot even when you have missing data. SageMaker Autopilot automatically fills in the missing data, provides statistical insights about columns in your dataset, and automatically extracts information from non-numeric columns, such as date and time information from timestamps. -
6
Reka
Reka
Our enterprise-grade multimodal assistant carefully designed with privacy, security, and efficiency in mind. We train Yasa to read text, images, videos, and tabular data, with more modalities to come. Use it to generate ideas for creative tasks, get answers to basic questions, or derive insights from your internal data. Generate, train, compress, or deploy on-premise with a few simple commands. Use our proprietary algorithms to personalize our model to your data and use cases. We design proprietary algorithms involving retrieval, fine-tuning, self-supervised instruction tuning, and reinforcement learning to tune our model on your datasets. -
7
DuckDB
DuckDB
Processing and storing tabular datasets, e.g. from CSV or Parquet files. Large result set transfer to client. Large client/server installations for centralized enterprise data warehousing. Writing to a single database from multiple concurrent processes. DuckDB is a relational database management system (RDBMS). That means it is a system for managing data stored in relations. A relation is essentially a mathematical term for a table. Each table is a named collection of rows. Each row of a given table has the same set of named columns, and each column is of a specific data type. Tables themselves are stored inside schemas, and a collection of schemas constitutes the entire database that you can access. -
8
MLBox
Axel ARONIO DE ROMBLAY
MLBox is a powerful Automated Machine Learning python library. It provides the following features fast reading and distributed data preprocessing/cleaning/formatting, highly robust feature selection and leak detection, accurate hyper-parameter optimization in high-dimensional space, state-of-the art predictive models for classification and regression (Deep Learning, Stacking, LightGBM), and prediction with models interpretation. MLBox main package contains 3 sub-packages: preprocessing, optimization and prediction. Each one of them are respectively aimed at reading and preprocessing data, testing or optimizing a wide range of learners and predicting the target on a test dataset. -
9
TimesFM-3
Google
TimesFM-3 is a state-of-the-art time series foundation model designed for highly accurate multivariate forecasting in a single forward pass. The 330 million parameter model is pre-trained on a real-world and synthetic time-series corpus comprising more than 1 trillion time points, building on the efficiency and zero-shot generalization of earlier TimesFM models. It can jointly predict multiple coevolving time series and capture dependencies that improve accuracy without task-specific fine-tuning. The model supports multiple targets with point and quantile forecasts, past covariates that are known only historically, and past-future dynamic covariates such as planned promotions, holidays, or weather forecasts. TimesFM-3 uses a decoder-only transformer architecture, processes contiguous data in patches of 32 time steps, and applies alternating causal temporal attention and full variate attention to combine patterns across time and related series. -
10
Webix Grid
Webix
Webix Grid is a standalone JavaScript DataGrid component (table/grid UI widget) that is optimized for high-performance, large-dataset scenarios, and is designed to be dropped into web applications where tabular data needs to be displayed, edited, filtered, sorted, etc. Key positioning points: Lightweight: you don’t have to bring in the full Webix UI library if you only need the grid. Focused on “just the grid” use-case rather than a full UI framework. Feature-rich, offering a wide set of capabilities for enterprise-style data apps: - Virtual scrolling - Frozen columns - Inline editing - Sorting - Filtering - Grouping - Column resizing - Column reordering - Multi-line headers - Row selection - Cell selection - Copy/paste - Excel export - PDF export - CSV export - Paging - Validation - Undo/redo - Drag and drop - Spanning cells - Context menus - Keyboard navigation - Conditional formatting - Column summaries - Custom rendering - Live updatesStarting Price: $749 per project -
11
PanGu-Σ
Huawei
Significant advancements in the field of natural language processing, understanding, and generation have been achieved through the expansion of large language models. This study introduces a system which utilizes Ascend 910 AI processors and the MindSpore framework to train a language model with over a trillion parameters, specifically 1.085T, named PanGu-{\Sigma}. This model, which builds upon the foundation laid by PanGu-{\alpha}, takes the traditionally dense Transformer model and transforms it into a sparse one using a concept known as Random Routed Experts (RRE). The model was efficiently trained on a dataset of 329 billion tokens using a technique called Expert Computation and Storage Separation (ECSS), leading to a 6.3-fold increase in training throughput via heterogeneous computing. Experimentation indicates that PanGu-{\Sigma} sets a new standard in zero-shot learning for various downstream Chinese NLP tasks. -
12
T5
Google
With T5, we propose reframing all NLP tasks into a unified text-to-text-format where the input and output are always text strings, in contrast to BERT-style models that can only output either a class label or a span of the input. Our text-to-text framework allows us to use the same model, loss function, and hyperparameters on any NLP task, including machine translation, document summarization, question answering, and classification tasks (e.g., sentiment analysis). We can even apply T5 to regression tasks by training it to predict the string representation of a number instead of the number itself. -
13
GPT-4
OpenAI
GPT-4 (Generative Pre-trained Transformer 4) is a large-scale unsupervised language model, yet to be released by OpenAI. GPT-4 is the successor to GPT-3 and part of the GPT-n series of natural language processing models, and was trained on a dataset of 45TB of text to produce human-like text generation and understanding capabilities. Unlike most other NLP models, GPT-4 does not require additional training data for specific tasks. Instead, it can generate text or answer questions using only its own internally generated context as input. GPT-4 has been shown to be able to perform a wide variety of tasks without any task specific training data such as translation, summarization, question answering, sentiment analysis and more.Starting Price: $0.0200 per 1000 tokens -
14
Odyssey-3
Odyssey
Odyssey-3 is a foundation world model and learned dynamical system that predicts how objects move and interact through space and how situations evolve over time. Implemented as an autoregressive diffusion transformer, it learns representations of physics, dynamics, and cause-and-effect from a broad dataset of visual observations and uses that knowledge to simulate environments and train policies for physical systems. Odyssey-3 can generate embodied environments from a prompt and predict in real time how they change as a person or agent acts or introduces events, with first-person and third-person navigation and independent camera movement. Its learned world knowledge can be adapted to different systems by training an action decoder or policy on paired observations and actions. With relatively small amounts of task-specific data, Odyssey-3 has been used to control robot arms, power humanoids, drive vehicles, and generate synchronized multi-sensor observations. -
15
Mistral OCR 4
Mistral AI
Mistral OCR 4 is a document extraction and understanding model built for enterprise search, RAG, domain-specific retrieval pipelines, and production-grade document intelligence. It extracts and structures content from a wide range of documents, moving beyond clean text and tables to return a structured representation of each page. Alongside extracted text, OCR 4 provides bounding boxes, typed-block classification, and inline confidence scores, helping downstream systems understand not only what the document says, but where each element sits, what role it plays, and how confident the model is in each region. Bounding boxes make in-context highlighting and reliable data pipelines possible, while block types and confidence scores support source-grounded citations, redactions, and human-in-the-loop verification. OCR 4 accepts common enterprise formats, including PDF, DOC, PPT, and OpenDocument, and supports 170 languages across 10 language groups.Starting Price: $2 per 1000 pages -
16
Intelligence Magic
Intelligence Magic
Say goodbye to messy research. Meet Intelligence Magic — the friendly AI that turns websites into organized, tabular insights in minutes! In 3 easy steps your everyday online research gets automated: 1. Describe Your Research: Tell us what data you need to gather across multiple websites. No coding required! 2. Define Your Columns: Specify exactly what columns you want in your spreadsheet - dates, prices, names, or any other data points. 3. Get Organized Results: Receive a clean, structured dataset ready for analysis or export.Starting Price: €50/month -
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
Jurassic-2
AI21
Announcing the launch of Jurassic-2, the latest generation of AI21 Studio’s foundation models, a game-changer in the field of AI, with top-tier quality and new capabilities. And that's not all, we're also releasing our task-specific APIs, with plug-and-play reading and writing capabilities that outperform competitors. Our focus at AI21 Studio is to help developers and businesses leverage reading and writing AI to build real-world products with tangible value. Today marks two important milestones with the release of Jurassic-2 and Task-Specific APIs, empowering you to bring generative AI to production. Jurassic-2 (or J2, as we like to call it) is the next generation of our foundation models with significant improvements in quality and new capabilities including zero-shot instruction-following, reduced latency, and multi-language support. Task-specific APIs provide developers with industry-leading APIs that perform specialized reading and writing tasks out-of-the-box.Starting Price: $29 per month -
19
VertiPaq Analyzer
SQLBI
VertiPaq Analyzer is useful to analyze VertiPaq storage structures for a data model in Power BI and Analysis Services Tabular. Added measures for segments and partitions: pageable, resident, refresh date, last access. Analysis Services provides many Dynamic Management Views (DMV) to collect information about memory used by a data model. For example, DISCOVER_OBJECT_MEMORY_USAGE is a DMV that provides information about all the objects in memory. You can use such a DMV also to monitor a Multidimensional instance of Analysis Services. Kasper de Jonge created a sample model (BISM Memory Report) that organizes this data in a hierarchical way, making it easy to find the most expensive databases, tables, and columns on a server. If you want to analyze a particular database, you probably want to look at more detailed information, which are available in other DMVs. -
20
thinkdeeply
Think Deeply
Discover from a variety of assets to jump-start your AI project. The AI hub provides a rich collection of artifacts that your project may need - industry AI starter kits, datasets, notebooks, pre-trained models, deployment-ready solutions & pipelines. Get access to the best resources from external parties, or created by your organization. Prepare and manage your data for model training. Collect, organize, tag, or select features, and prepare datasets for training with simple drag and drop UI. Collaborate with multiple team members to tag large datasets. Implement a quality control process to ensure dataset quality. Build models with simple clicks using the model wizards. No data science knowledge required. The system selects the best models for the problem and optimizes their training parameters. Advanced users, however, can fine-tune the models and their hyper-parameters. One-click deployment to production inference enviornments. -
21
Llama 3.1
Meta
The open source AI model you can fine-tune, distill and deploy anywhere. Our latest instruction-tuned model is available in 8B, 70B and 405B versions. Using our open ecosystem, build faster with a selection of differentiated product offerings to support your use cases. Choose from real-time inference or batch inference services. Download model weights to further optimize cost per token. Adapt for your application, improve with synthetic data and deploy on-prem or in the cloud. Use Llama system components and extend the model using zero shot tool use and RAG to build agentic behaviors. Leverage 405B high quality data to improve specialized models for specific use cases.Starting Price: Free -
22
Mistral NeMo
Mistral AI
Mistral NeMo, our new best small model. A state-of-the-art 12B model with 128k context length, and released under the Apache 2.0 license. Mistral NeMo is a 12B model built in collaboration with NVIDIA. Mistral NeMo offers a large context window of up to 128k tokens. Its reasoning, world knowledge, and coding accuracy are state-of-the-art in its size category. As it relies on standard architecture, Mistral NeMo is easy to use and a drop-in replacement in any system using Mistral 7B. We have released pre-trained base and instruction-tuned checkpoints under the Apache 2.0 license to promote adoption for researchers and enterprises. Mistral NeMo was trained with quantization awareness, enabling FP8 inference without any performance loss. The model is designed for global, multilingual applications. It is trained on function calling and has a large context window. Compared to Mistral 7B, it is much better at following precise instructions, reasoning, and handling multi-turn conversations.Starting Price: Free -
23
Nixtla
Nixtla
Nixtla is a platform for time-series forecasting and anomaly detection built around its flagship model TimeGPT, described as the first generative AI foundation model for time-series data. It was trained on over 100 billion data points spanning domains such as retail, energy, finance, IoT, healthcare, weather, web traffic, and more, allowing it to make accurate zero-shot predictions across a wide variety of use cases. With just a few lines of code (e.g., via their Python SDK), users can supply historical data and immediately generate forecasts or detect anomalies, even for irregular or sparse time series, and without needing to build or train models from scratch. TimeGPT supports advanced features like handling exogenous variables (e.g., events, prices), forecasting multiple time-series at once, custom loss functions, cross-validation, prediction intervals, and model fine-tuning on bespoke datasets.Starting Price: Free -
24
Goku
ByteDance
The Goku AI model, developed by ByteDance, is an open source advanced artificial intelligence system designed to generate high-quality video content based on given prompts. It utilizes deep learning techniques to create stunning visuals and animations, particularly focused on producing realistic, character-driven scenes. By leveraging state-of-the-art models and a vast dataset, Goku AI allows users to create custom video clips with incredible accuracy, transforming text-based input into compelling and immersive visual experiences. The model is particularly adept at producing dynamic characters, especially in the context of popular anime and action scenes, offering creators a unique tool for video production and digital content creation.Starting Price: Free -
25
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 -
26
Qwen3.8-Flash-Next
Alibaba
Qwen3.8-Flash-Next is an open-weight multimodal Mixture-of-Experts model and an early preview of the architecture planned for Qwen4. It systematically upgrades attention, residual connections, embeddings, and optimization to improve capability, computational efficiency, model capacity, and training stability. Its hybrid architecture combines Gated DeltaNet, which efficiently compresses historical information, with Qwen Sparse Attention, which selects important context at the micro-block level to reduce attention and indexing costs on long sequences. Gated Residual widens the residual stream into four branches and dynamically controls information flow across layers, while N-gram Embedding adds large-scale local-pattern memory with very little extra per-token computation and can be offloaded to host memory. The model uses a 125B-parameter main network plus 51B N-gram embedding parameters, while activating only 6B parameters per token.Starting Price: $2 per 1M (input) -
27
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 -
28
Qvu Data Service
rbtdesign
Qvu Data Service is an ad-hoc query and api data service design tool that allows users to create and save query designs in a user-friendly, web-based UI. Qvu Data Service provides REST API endpoints for users and applications to execute saved query documents and return results in tabular or JSON formatted result sets. Qvu Data Service provides role-based data source, table column and document group access control and supports both Basic and OIDC authentication.Starting Price: $0 -
29
Pivot Desktop
Pivot Inc.
Pivot Desktop is a Windows desktop workspace for analysts, researchers, and students to prepare, explore, visualize, and model tabular data. Users build editable, repeatable workflows by arranging data-cleaning, transformation, statistical-analysis, charting, and machine-learning steps in a visual pipeline, then save sessions or export results. Pivot opens CSV, Excel, Parquet, and JSON files and can connect to PostgreSQL, MySQL, SQL Server, or SQLite. It includes interactive charts, statistical tests, regression and other models, plus a local AI assistant that turns natural-language requests into pipeline steps. Data and the AI assistant run on the user's machine. An optional Model Context Protocol server lets external agents build workflows programmatically. -
30
Lightning Rod
Lightning Rod
Lightning Rod is an AI platform designed to transform messy, unstructured real-world data into verified, production-ready training datasets and domain-specific AI models without requiring manual labeling. It enables users to generate high-quality, citable question–answer pairs from sources such as news articles, financial filings, and internal documents, turning raw historical data into structured datasets that can be used for supervised fine-tuning or reinforcement learning. It operates through an agent-driven workflow where users describe their goal, and the system automatically gathers sources, generates questions, resolves outcomes based on real-world events, and adds contextual grounding before training a model. A key innovation is its “future-as-label” methodology, which uses actual outcomes as training signals, allowing AI systems to learn directly from real-world results at scale instead of relying on synthetic or manually annotated data. -
31
DataGen
DataGen
DataGen is a leading AI platform specializing in synthetic data generation and custom generative AI models for machine learning projects. Their flagship product, SynthEngyne, supports multi-format data generation including text, images, tabular, and time-series data, ensuring privacy-compliant, high-quality training datasets. The platform offers scalable, real-time processing and advanced quality controls like deduplication to maintain dataset fidelity. DataGen also provides professional AI development services such as model deployment, fine-tuning, synthetic data consulting, and intelligent automation systems. With flexible pricing plans ranging from free tiers for individuals to custom enterprise solutions, DataGen caters to a wide range of users. Their solutions serve diverse industries including healthcare, finance, automotive, and retail. -
32
Port
Port IO
Port is a platform for building no-code, holistic, Internal Developer Portals. Port's software catalog covers microservices, resources, custom assets and fits any data model, with in-context maturity scorecards. Its portals support any developer self-service action and workflow automation. -
33
Tülu 3
Ai2
Tülu 3 is an advanced instruction-following language model developed by the Allen Institute for AI (Ai2), designed to enhance capabilities in areas such as knowledge, reasoning, mathematics, coding, and safety. Built upon the Llama 3 Base, Tülu 3 employs a comprehensive four-stage post-training process: meticulous prompt curation and synthesis, supervised fine-tuning on a diverse set of prompts and completions, preference tuning using both off- and on-policy data, and a novel reinforcement learning approach to bolster specific skills with verifiable rewards. This open-source model distinguishes itself by providing full transparency, including access to training data, code, and evaluation tools, thereby closing the performance gap between open and proprietary fine-tuning methods. Evaluations indicate that Tülu 3 outperforms other open-weight models of similar size, such as Llama 3.1-Instruct and Qwen2.5-Instruct, across various benchmarks.Starting Price: Free -
34
GPT-J
EleutherAI
GPT-J is a cutting-edge language model created by the research organization EleutherAI. In terms of performance, GPT-J exhibits a level of proficiency comparable to that of OpenAI's renowned GPT-3 model in a range of zero-shot tasks. Notably, GPT-J has demonstrated the ability to surpass GPT-3 in tasks related to generating code. The latest iteration of this language model, known as GPT-J-6B, is built upon a linguistic dataset referred to as The Pile. This dataset, which is publicly available, encompasses a substantial volume of 825 gibibytes of language data, organized into 22 distinct subsets. While GPT-J shares certain capabilities with ChatGPT, it is important to note that GPT-J is not designed to operate as a chatbot; rather, its primary function is to predict text. In a significant development in March 2023, Databricks introduced Dolly, a model that follows instructions and is licensed under Apache.Starting Price: Free -
35
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 -
36
Kimi K2
Moonshot AI
Kimi K2 is a state-of-the-art open source large language model series built on a mixture-of-experts (MoE) architecture, featuring 1 trillion total parameters and 32 billion activated parameters for task-specific efficiency. Trained with the Muon optimizer on over 15.5 trillion tokens and stabilized by MuonClip’s attention-logit clamping, it delivers exceptional performance in frontier knowledge, reasoning, mathematics, coding, and general agentic workflows. Moonshot AI provides two variants, Kimi-K2-Base for research-level fine-tuning and Kimi-K2-Instruct pre-trained for immediate chat and tool-driven interactions, enabling both custom development and drop-in agentic capabilities. Benchmarks show it outperforms leading open source peers and rivals top proprietary models in coding tasks and complex task breakdowns, while its 128 K-token context length, tool-calling API compatibility, and support for industry-standard inference engines.Starting Price: Free -
37
Grok 4.8
SpaceXSI
Grok 4.8 is an upcoming AI model from SpaceXSI expected to advance the Grok family in reasoning, coding, agentic workflows, and professional knowledge work. Elon Musk has described the model as having approximately 2.5 trillion parameters and being trained using a new C++ software stack. The model is expected to complete its initial training before entering reinforcement learning, with final capabilities and performance still subject to change. Grok 4.8 is anticipated to build on Grok 4.7’s strengths in software development, tool calling, configurable reasoning, multimodal input, and long-running agentic tasks. SpaceXSI has not yet released official benchmarks, pricing, context-window specifications, API identifiers, or a public launch date for Grok 4.8. The model is expected to target developers, researchers, enterprises, and advanced AI users who need high-capability reasoning and autonomous task execution. -
38
Qwen 4
Alibaba
Qwen 4 is Alibaba’s next-generation Qwen foundation model, announced on September 22, 2026, and currently in training. It will follow the Qwen3.8 generation as part of Alibaba’s broader roadmap for increasingly capable foundation and agentic AI models. Alibaba has not yet published Qwen 4’s parameter count, architecture, benchmark results, context window, pricing, release date, or availability details. The announcement places Qwen 4 within a research strategy focused on large-scale model training, agentic reinforcement learning, multimodal intelligence, and recursive self-improvement. Alibaba separately said that later Qwen 4.5 and Qwen 5 models are expected to scale into the 5-to-10-trillion-parameter range, but that figure was not attributed specifically to Qwen 4. Because Qwen 4 has not yet been released, detailed performance comparisons and production capabilities remain unconfirmed. -
39
Codestral
Mistral AI
We introduce Codestral, our first-ever code model. Codestral is an open-weight generative AI model explicitly designed for code generation tasks. It helps developers write and interact with code through a shared instruction and completion API endpoint. As it masters code and English, it can be used to design advanced AI applications for software developers. Codestral is trained on a diverse dataset of 80+ programming languages, including the most popular ones, such as Python, Java, C, C++, JavaScript, and Bash. It also performs well on more specific ones like Swift and Fortran. This broad language base ensures Codestral can assist developers in various coding environments and projects.Starting Price: Free -
40
GLM-5
Z.ai
GLM-5 is Z.ai’s latest large language model built for complex systems engineering and long-horizon agentic tasks. It scales significantly beyond GLM-4.5, increasing total parameters and training data while integrating DeepSeek Sparse Attention to reduce deployment costs without sacrificing long-context capacity. The model combines enhanced pre-training with a new asynchronous reinforcement learning infrastructure called slime, improving training efficiency and post-training refinement. GLM-5 achieves best-in-class performance among open-source models across reasoning, coding, and agent benchmarks, narrowing the gap with leading frontier models. It ranks highly on evaluations such as Vending Bench 2, demonstrating strong long-term planning and operational capabilities. The model is open-sourced under the MIT License.Starting Price: Free -
41
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. -
42
Amundsen
Amundsen
Discover & trust data for your analysis and models. Be more productive by breaking silos. Get immediate context into the data and see how others are using it. Search for data within your organization by a simple text search. A PageRank-inspired search algorithm recommends results based on names, descriptions, tags, and querying/viewing activity on the table/dashboard. Build trust in data using automated and curated metadata, descriptions of tables and columns, other frequent users, when the table was last updated, statistics, a preview of the data if permitted, etc. Easy triage by linking the ETL job and code that generated the data. Update tables and columns with descriptions, reduce unnecessary back and forth about which table to use and what a column contains. See what data fellow co-workers frequently use, own or have bookmarked. Learn what most common queries for a table look like by seeing dashboards built on a given table. -
43
OpenEuroLLM
OpenEuroLLM
OpenEuroLLM is a collaborative initiative among Europe's leading AI companies and research institutions to develop a series of open-source foundation models for transparent AI in Europe. The project emphasizes transparency by openly sharing data, documentation, training, testing code, and evaluation metrics, fostering community involvement. It ensures compliance with EU regulations, aiming to provide performant large language models that align with European standards. A key focus is on linguistic and cultural diversity, extending multilingual capabilities to encompass all EU official languages and beyond. The initiative seeks to enhance access to foundational models ready for fine-tuning across various applications, expand evaluation results in multiple languages, and increase the availability of training datasets and benchmarks. Transparency is maintained throughout the training processes by sharing tools, methodologies, and intermediate results. -
44
Mistral Large 2
Mistral AI
Mistral AI has launched the Mistral Large 2, an advanced AI model designed to excel in code generation, multilingual capabilities, and complex reasoning tasks. The model features a 128k context window, supporting dozens of languages including English, French, Spanish, and Arabic, as well as over 80 programming languages. Mistral Large 2 is tailored for high-throughput single-node inference, making it ideal for large-context applications. Its improved performance on benchmarks like MMLU and its enhanced code generation and reasoning abilities ensure accuracy and efficiency. The model also incorporates better function calling and retrieval, supporting complex business applications.Starting Price: Free -
45
Cleanlab
Cleanlab
Cleanlab Studio handles the entire data quality and data-centric AI pipeline in a single framework for analytics and machine learning tasks. Automated pipeline does all ML for you: data preprocessing, foundation model fine-tuning, hyperparameter tuning, and model selection. ML models are used to diagnose data issues, and then can be re-trained on your corrected dataset with one click. Explore the entire heatmap of suggested corrections for all classes in your dataset. Cleanlab Studio provides all of this information and more for free as soon as you upload your dataset. Cleanlab Studio comes pre-loaded with several demo datasets and projects, so you can check those out in your account after signing in. -
46
VideoPoet
Google
VideoPoet is a simple modeling method that can convert any autoregressive language model or large language model (LLM) into a high-quality video generator. It contains a few simple components. An autoregressive language model learns across video, image, audio, and text modalities to autoregressively predict the next video or audio token in the sequence. A mixture of multimodal generative learning objectives are introduced into the LLM training framework, including text-to-video, text-to-image, image-to-video, video frame continuation, video inpainting and outpainting, video stylization, and video-to-audio. Furthermore, such tasks can be composed together for additional zero-shot capabilities. This simple recipe shows that language models can synthesize and edit videos with a high degree of temporal consistency. -
47
GPT-5.2 Thinking
OpenAI
GPT-5.2 Thinking is the highest-capability configuration in OpenAI’s GPT-5.2 model family, engineered for deep, expert-level reasoning, complex task execution, and advanced problem solving across long contexts and professional domains. Built on the foundational GPT-5.2 architecture with improvements in grounding, stability, and reasoning quality, this variant applies more compute and reasoning effort to generate responses that are more accurate, structured, and contextually rich when handling highly intricate workflows, multi-step analysis, and domain-specific challenges. GPT-5.2 Thinking excels at tasks that require sustained logical coherence, such as detailed research synthesis, advanced coding and debugging, complex data interpretation, strategic planning, and sophisticated technical writing, and it outperforms lighter variants on benchmarks that test professional skills and deep comprehension. -
48
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 -
49
GPT-5.4 Pro
OpenAI
GPT-5.4 Pro is an advanced AI model developed by OpenAI to deliver high-performance capabilities for professional and complex tasks. It combines improvements in reasoning, coding, and agent-based workflows into a single unified system. The model is designed to work efficiently across professional tools such as spreadsheets, presentations, documents, and development environments. GPT-5.4 Pro also includes native computer-use capabilities, enabling AI agents to interact with software, websites, and operating systems to complete tasks. With support for up to one million tokens of context, it can manage long workflows and large datasets more effectively than previous models. The model also improves tool usage, allowing it to search for and select the right tools during multi-step processes. By delivering more accurate outputs with fewer tokens, GPT-5.4 Pro helps professionals complete complex work faster and more efficiently. -
50
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
