![](\"figures/granite-code-models-3x-v4.png\")
\n\n\n
\n:books: Paper | :hugs: HuggingFace Collection | \n:speech_balloon: Discussions Page \n
\n\n---\n## Introduction to Granite Code Models\nWe introduce the Granite series of decoder-only code models for code generative tasks (e.g., fixing bugs, explaining code, documenting code), trained with code written in 116 programming languages. A comprehensive evaluation of the Granite Code model family on diverse tasks demonstrates that our models consistently reach state-of-the-art performance among available open source code LLMs. \n\nThe key advantages of Granite Code models include:\n* All-rounder Code LLM: Granite Code models achieve competitive or state-of-the-art performance on different kinds of code-related tasks, including code generation, explanation, fixing, editing, translation, and more. Demonstrating their ability to solve diverse coding tasks.\n* Trustworthy Enterprise-Grade LLM: All our models are trained on license-permissible data collected following [IBM's AI Ethics principles](https://www.ibm.com/impact/ai-ethics) and guided by IBM’s Corporate Legal team for trustworthy enterprise usage. We release all our Granite Code models under an [Apache 2.0 license](https://www.apache.org/licenses/LICENSE-2.0) license for research and commercial use.\n\nThe family of **Granite Code Models** comes in two main variants:\n\n* Granite Code Base Models: base foundational models designed for code-related tasks (e.g., code repair, code explanation, code synthesis).\n* Granite Code Instruct Models: instruction following models finetuned using a combination of Git com"},{"ref":"P3","kind":"page","title":"ibm-granite/discussions repository metadata","date":"2026-06-11T04:19:58.324034+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/discussions","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/discussions\n\nStars: 1\n\nForks: 0\n\nOpen issues: 1\n\nCreated: 2024-04-29T08:50:33Z\n\nPushed: 2025-06-25T20:35:03Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\nThis repository hosts [discussions](https://github.com/ibm-granite/discussions/discussions) for this organization.\nDo NOT delete, rename or modify this repo."},{"ref":"P4","kind":"page","title":"ibm-granite/watsonx-code-assistant-individual repository metadata","date":"2026-06-11T04:19:58.280544+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/watsonx-code-assistant-individual","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/watsonx-code-assistant-individual\n\nDescription: For individual users, watsonx Code Assistant can access a local IBM Granite model \n\nLicense: Apache-2.0\n\nStars: 37\n\nForks: 6\n\nOpen issues: 7\n\nCreated: 2024-08-30T21:48:40Z\n\nPushed: 2025-06-25T20:35:29Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n# License\n- The license for IBM watsonx Code Assistant Individual can be found in the [product-licenses](./product-licenses/) folder in this repository.\n\n- You can use this repository to add issues for watsonx Code Assistant Individual. The license for issues, discussion, and any files or samples shared in issues can be found in the [LICENSE](./LICENSE) file.\n\n# IBM watsonx Code Assistant Individual\n\n## Features\n\nWatsonx Code Assistant Individual is an innovative, lightweight AI coding companion built for IBM’s state-of-the-art Granite large language models. This companion offers robust, contextually aware AI coding assistance for popular programming languages such as C, C++, Go, Java, JavaScript, Python, and TypeScript. Seamlessly integrated into Visual Studio Code, watsonx Code Assistant Individual accelerates development productivity and simplifies coding tasks by providing powerful AI support hosted locally on the developer’s laptop or workstation using Ollama.\n\n### Chat with code models\n\n- Chat with an IBM Granite code model to create code, and ask general programming questions.\n- Use the chat to explain and extend existing code from your workspace.\n\n\n\n### Code completion\n\nComplete the line that you're currently typing:\n\n![\"Single-line](\"https://github.com/ibm-granite/watsonx-code-assistant-individual/raw/HEAD/images/Single-line.gif\")
\n\nAnd even full methods and functions:\n\n![\"Multi-line](\"https://github.com/ibm-granite/watsonx-code-assistant-individual/raw/HEAD/images/Multi-line.gif\")
\n\n### Turn comments into code\n\nCreate a comment that describes a "},{"ref":"P5","kind":"page","title":"ibm-granite/.github repository metadata","date":"2026-06-11T04:19:58.277565+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/.github","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/.github\n\nStars: 2\n\nForks: 0\n\nOpen issues: 1\n\nCreated: 2024-04-29T01:52:19Z\n\nPushed: 2026-04-30T18:30:11Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\nThis repository hosts information about the IBM-Granite organization."},{"ref":"P6","kind":"page","title":"ibm-granite/granite-geospatial-biomass repository metadata","date":"2026-06-11T04:19:58.275826+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/granite-geospatial-biomass","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/granite-geospatial-biomass\n\nDescription: A geospatial model for Above Ground Biomass\n\nLanguage: Jupyter Notebook\n\nLicense: Apache-2.0\n\nStars: 28\n\nForks: 10\n\nOpen issues: 4\n\nCreated: 2024-06-27T18:22:34Z\n\nPushed: 2024-09-18T17:32:51Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n# A Fine-Tuned Geospatial Foundation Model for Biomass\n\nPublic notebooks and utilities for working with the fine-tuned granite.geospatial.biomass model\n\n## Initial Setup\n\nFirst clone the repository:\n\n```sh\ngit clone https://github.com/ibm-granite/granite-geospatial-biomass.git\n```\n\nThen run `pip install -e .` to install all necessary dependencies to run the notebook\n\n## 📕 Notebook\n\nCheck out the [Getting Started Notebook!](https://github.com/ibm-granite/granite-geospatial-biomass/blob/main/notebooks/agb_getting_started.ipynb)\n\n## IBM Public Repository Disclosure\n\nAll content in this repository including code has been provided by IBM under the associated open source software license and IBM is under no obligation to provide enhancements, updates, or support. IBM developers produced this code as an open source project (not as an IBM product), and IBM makes no assertions as to the level of quality nor security, and will not be maintaining this code going forward."},{"ref":"P7","kind":"page","title":"ibm-granite/granite-3.0-language-models repository metadata","date":"2026-06-11T04:19:57.478038+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/granite-3.0-language-models","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/granite-3.0-language-models\n\nLicense: Apache-2.0\n\nStars: 270\n\nForks: 28\n\nOpen issues: 4\n\nCreated: 2024-10-18T20:01:15Z\n\nPushed: 2025-06-25T20:35:53Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n
\n![](\"https://github.com/ibm-granite/granite-3.0-language-models/blob/main/figures/granite-3_0-language-models-3x-v1.png\")
\n
\n:books: Paper | :hugs: HuggingFace Collection | \n:speech_balloon: Discussions Page \n
\n\n---\n## Introduction to Granite 3.0 Language Models\nGranite 3.0 language models are a new set of lightweight state-of-the-art, open foundation models that natively support multilinguality, coding, reasoning, and tool usage, including the potential to be run on constrained compute resources. All the models are publicly released under an Apache 2.0 license for both research and commercial use. The models' data curation and training procedure were designed for enterprise usage and customization in mind, with a process that evaluates datasets for governance, risk and compliance (GRC) criteria, in addition to IBM's standard data clearance process and document quality checks.\n\nGranite 3.0 includes 4 different models of varying sizes:\n- Dense Models: 2B and 8B parameter models, trained on 12 trillion tokens in total.\n- Mixture-of-Expert (MoE) Models: Sparse 1B and 3B MoE models, with 400M and 800M activated parameters respectively, trained on 10 trillion tokens in total.\n\nAccordingly, these options provide a range of models with different compute requirements to choose from, with appropriate trade-offs with their performance on downstream tasks. At each scale, we release a base model — checkpoints of models after pretraining, as well as instruct checkpoints — models finetuned for dialogue, instruction-following, helpfulness, and safety.\n\n## Data Collection\nGranite 3.0 language models are trained using data from various so"},{"ref":"P8","kind":"page","title":"ibm-granite/granite-geospatial-land-surface-temperature repository metadata","date":"2026-06-11T04:19:57.434312+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/granite-geospatial-land-surface-temperature","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/granite-geospatial-land-surface-temperature\n\nLanguage: Jupyter Notebook\n\nLicense: Apache-2.0\n\nStars: 14\n\nForks: 7\n\nOpen issues: 1\n\nCreated: 2024-11-05T09:14:49Z\n\nPushed: 2025-11-01T19:36:35Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n# granite-geospatial-land-surface-temperature\n\nPublic notebooks and utilities for working with the fine-tuned granite.geospatial.land-surface-temperature model\n\n## Initial Setup\n\nFirst clone the repository:\n\n```sh\ngit clone https://github.com/ibm-granite/granite-geospatial-land-surface-temperature\n```\n\nThen run `pip install -e .` to install all necessary dependencies to run the notebooks\n\n## 📕 Notebooks\n\n1. Check out the [Getting Started Notebook!](https://github.com/ibm-granite/granite-geospatial-land-surface-temperature/blob/main/notebooks/1_getting_started.ipynb).\n2. For Tweening (Temporal Gap-Filling) check out the [Introduction to LST Tweening Notebook!](https://github.com/ibm-granite/granite-geospatial-land-surface-temperature/blob/main/notebooks/2_introduction_to_LST_Tweening.ipynb) for a tutorial on how to implement Tweening and the [Tweening Data Preparation Notebook!](https://github.com/ibm-granite/granite-geospatial-land-surface-temperature/blob/main/notebooks/3_tweening_data_preparation.ipynb) for a tutorial on preparing the data for Tweening.\n3. For data download and data pre-processing to create your own dataset check out the [Download Notebook!](https://github.com/ibm-granite/granite-geospatial-land-surface-temperature/blob/main/notebooks/4_download_data.ipynb) and the [Preprocessing Notebook!](https://github.com/ibm-granite/granite-geospatial-land-surface-temperature/blob/main/notebooks/5_preprocess_data.ipynb).\n\n## IBM Public Repository Disclosure\n\nAll content in this repository including code has been provided by IBM under the associated open source software license and IBM is under no obligation to provide enhancements, updates, or support. IBM developers produced this code as an open source project (not as an IBM product), and IBM makes no assertions as to the level of quality nor security, and will not be maintaining this code going forward."},{"ref":"P9","kind":"page","title":"ibm-granite/granite-guardian repository metadata","date":"2026-06-11T04:19:57.431112+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/granite-guardian","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/granite-guardian\n\nDescription: The Granite Guardian models are designed to detect risks in prompts and responses.\n\nLanguage: Jupyter Notebook\n\nLicense: Apache-2.0\n\nStars: 152\n\nForks: 17\n\nOpen issues: 9\n\nCreated: 2024-10-11T14:28:45Z\n\nPushed: 2026-05-05T01:44:52Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n# Granite Guardian\n\n📌 What's New?\n\n✨ **April 2026**: [Granite-Guardian-4.1-8B](https://huggingface.co/ibm-granite/granite-guardian-4.1-8b) introduces improved Bring Your Own Criteria (BYOC) support, enabling users to define arbitrary judging criteria beyond the pre-baked safety and hallucination detectors. The model can now faithfully evaluate complex, multi-part requirements such as formatting rules, length constraints, and domain-specific instructions.\n\n✨ **Sept 2025**: 🏆 [Granite-Guardian-3.3](https://huggingface.co/ibm-granite/granite-guardian-3.3-8b) has has secured the 3rd position on the [LLM‑AggreFact benchmark](https://llm-aggrefact.github.io/), a comprehensive fact‑checking benchmark that consolidates 11 datasets on grounded factuality. Granite Guardian 3.3 8B also holds the #1 position on the [REVEAL benchmark](https://research.google/pubs/a-chain-of-thought-is-as-strong-as-its-weakest-link-a-benchmark-for-verifiers-of-reasoning-chains/) (a dataset that evaluates the correctness of reasoning chains generated by LLMs) which is one of the 11 dimensions in LLM-AggreFact. Additionally, while our Granite Guardian model is only 8B in parameter size, it outperforms much larger models such as gpt-4o and Mistral Large 2 on this benchmark.\n\n✨ **Sept 2025**: Two new LoRA adapters for [multi-risk detection](https://huggingface.co/ibm-granite/granite-guardian-3.2-5b-lora-harm-categories) and [harm-correction](https://huggingface.co/ibm-granite/granite-guardian-3.2-5b-lora-harm-correction) are live!\n\n✨ **Aug 2025**: [Granite-Guardian-3.3](https://huggingface.co/ibm-granite/granite-guardian-3.3-8b) is live! 🤖 New hybrid thinking mode for better reasoning and improved bring-your-own-criteria functionality.\n\n✨ Feb 2025: Granite-Guardian-3.2 is out! ⚙️ Adds two new model sizes, verbalized confidence, and two new risks. Updated notebooks inc"},{"ref":"P10","kind":"page","title":"ibm-granite/granite-geospatial-canopyheight repository metadata","date":"2026-06-11T04:19:57.39234+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/granite-geospatial-canopyheight","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/granite-geospatial-canopyheight\n\nLanguage: Jupyter Notebook\n\nLicense: Apache-2.0\n\nStars: 5\n\nForks: 3\n\nOpen issues: 2\n\nCreated: 2024-11-05T08:44:30Z\n\nPushed: 2024-11-08T14:37:35Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n# A Fine-Tuned Geospatial Foundation Model for Canopy Height Model\n\nPublic notebooks and utilities for working with the fine-tuned granite-geospatial-canopyheight model which you can find on [Hugging Face](https://huggingface.co/ibm-granite/granite-geospatial-canopyheight).\n\n## Initial Setup\n\nFirst clone the repository:\n\n```sh\ngit clone https://github.com/ibm-granite/granite-geospatial-canopyheight.git\n```\n\nThen run `pip install -e .` to install all necessary dependencies to run the notebook\n\n## 📕 Notebook\n\nCheck out the [Getting Started Notebook!](https://github.com/ibm-granite/granite-geospatial-canopyheight/blob/main/notebooks/canopyheight_getting_started.ipynb)\n\n## IBM Public Repository Disclosure\n\nAll content in this repository including code has been provided by IBM under the associated open source software license and IBM is under no obligation to provide enhancements, updates, or support. IBM developers produced this code as an open source project (not as an IBM product), and IBM makes no assertions as to the level of quality nor security, and will not be maintaining this code going forward."},{"ref":"P11","kind":"page","title":"ibm-granite/geospatial repository metadata","date":"2026-06-11T04:19:57.364364+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/geospatial","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/geospatial\n\nLanguage: Jupyter Notebook\n\nStars: 15\n\nForks: 10\n\nOpen issues: 3\n\nCreated: 2024-12-02T15:04:32Z\n\nPushed: 2026-01-19T10:05:51Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\nThis repository contains code related to the [granite geospatial](https://huggingface.co/collections/ibm-granite/granite-geospatial-models-667dacfed21bdcf60a8bc982) models IBM opensourced and released on Huggingface."},{"ref":"P12","kind":"page","title":"ibm-granite/granite-io repository metadata","date":"2026-06-11T04:19:56.595223+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/granite-io","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/granite-io\n\nDescription: ⚠️ Deprecated: This library's functionality has been rolled into Mellea (https://github.com/generative-computing/mellea)\n\nLanguage: Python\n\nLicense: Apache-2.0\n\nStars: 57\n\nForks: 30\n\nOpen issues: 25\n\nCreated: 2025-02-20T17:29:55Z\n\nPushed: 2026-05-18T17:34:16Z\n\nDefault branch: main\n\nFork: no\n\nArchived: yes\n\nREADME:\n
![\"IBM](\"./mkdocs/docs/img/ibm_logo_rebus.png\"){kind=logo}
\n\n# Granite IO Processing\n\n[](https://pypi.org/project/granite-io/)\n[](https://pypi.org/project/granite-io/)\n[](LICENSE)\n[](https://github.com/ibm-granite/granite-io/stargazers)\n[](./mkdocs/)\n[](https://pypi.org/project/granite-io/)\n\n**🚀 A powerful framework for extending IBM Granite model functionality through input/output processing**\n\n\n![](\"https://github.com/ibm-granite/granite-3.1-language-models/blob/main/figures/granite-3_1-language-models-3x-v1.png\")
\n
\n:books: Paper (comming soon) | :hugs: HuggingFace Collection | \n:speech_balloon: Discussions Page | 📘 IBM Granite Docs\n
\n\n---\n## Introduction to Granite 3.1 Language Models\nGranite 3.1 language models are lightweight, state-of-the-art, open foundation models that natively support multilinguality, coding, reasoning, and tool usage, including the potential to be run on constrained compute resources. All the models are publicly released under an Apache 2.0 license for both research and commercial use. The models' data curation and training procedure were designed for enterprise usage and customization, with a process that evaluates datasets for governance, risk and compliance (GRC) criteria, in addition to IBM's standard data clearance process and document quality checks.\n\nGranite 3.1 language models extend the context length of Granite 3.0 language models from 4K to 128K using a progressive training strategy by increasing the supported context length in increments while adjusting RoPE theta until the models successfully adapt to the desired length of 128K. This long-context pre-training stage was performed using approximately 500B tokens. Moreover, Granite 3.1 instruction models provide an improved developer experience for function-calling and RAG generation tasks.\n\nGranite 3.1 models come in 4 varying sizes and 2 architectures:\n- Dense Models: 2B and 8B parameter models, trained on 12 trillion tokens in total.\n- Mixture-of-Exper"},{"ref":"P14","kind":"page","title":"ibm-granite/granite-embedding-models repository metadata","date":"2026-06-11T04:19:56.491803+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/granite-embedding-models","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/granite-embedding-models\n\nLanguage: Jupyter Notebook\n\nLicense: Apache-2.0\n\nStars: 71\n\nForks: 4\n\nOpen issues: 4\n\nCreated: 2024-12-10T16:14:10Z\n\nPushed: 2026-05-14T02:40:11Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n---\n## Introduction to Granite Embedding Models\n\n- **Huggingface Repository:** [ibm-granite/granite-embedding-models](https://huggingface.co/collections/ibm-granite/granite-embedding-models-6750b30c802c1926a35550bb)\n- **Documentation**: [Granite Docs](https://www.ibm.com/granite/docs/models/embedding/)\n- **Granite Community**: [ibm-granite-community](https://github.com/ibm-granite-community)\n\nThe Granite Embedding collection delivers innovative sentence-transformer models purpose-built for retrieval-based applications. Featuring a bi-encoder architecture, these models generate high-quality embeddings for textual inputs such as queries, passages, and documents, enabling seamless comparison through cosine similarity. Built using retrieval oriented pretraining, contrastive finetuning, knowledge distillation, and model merging, the Granite Embedding lineup is optimized to ensure strong alignment between query and passage embeddings. \n\nBuilt on a foundation of carefully curated, permissibly licensed public datasets, the Granite Embedding models set a high standard for performance, maintaining competitive scores not only on academic benchmarks such as BEIR, but also out-perfoming models of the same size on many enterprise use cases. Developed to meet enterprise-grade expectations, they are crafted transparently in accordance with IBM's AI Ethics principles and offered under the Apache 2.0 license for both research and commercial innovation. \n\nThe R2 models show strong performance across standard and IBM-built information retrieval benchmarks (BEIR, ClapNQ),\ncode retrieval (COIR), long-document search benchmarks (MLDR, LongEmbed), conversational multi-turn (MTRAG),\ntable retrieval (NQTables, OTT-QA, AIT-QA, MultiHierTT, OpenWikiTables), and on many enterprise use cases. For details on improvements over R1, see [What's New in R2](whats_new_in_r2.md).\n\nThese models use a bi-encoder architecture to generate high-quality embeddings from"},{"ref":"P15","kind":"page","title":"ibm-granite/granite-vision-models repository metadata","date":"2026-06-11T04:19:56.481669+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/granite-vision-models","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/granite-vision-models\n\nLanguage: Jupyter Notebook\n\nLicense: Apache-2.0\n\nStars: 46\n\nForks: 10\n\nOpen issues: 5\n\nCreated: 2025-02-17T21:42:13Z\n\nPushed: 2026-04-29T22:17:48Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n
\n:books: Granite Vision Paper | \n:bar_chart: ChartNet CVPR 2026 Paper | \n:hugs: HuggingFace Collection | \n:speech_balloon: Discussions Page \n
\n\n# Granite Vision Models\n\nGranite Vision is a family of multimodal vision‑language models designed to support enterprise‑grade document understanding tasks, including charts, tables, key‑value extraction, and structured image‑to‑text generation. \nThis repository provides documentation, examples, and pointers to available model releases and datasets.\n\n---\n\n## 🚀 Latest Release: Granite‑Vision-4.1‑4B \n\n[Granite‑Vision-4.1‑4B](https://huggingface.co/ibm-granite/granite-vision-4.1-4b) is a vision‑language model tailored for **enterprise document data extraction**, delivered as a **LoRA adapter** on top of Granite-4.1-3B.\n\nIt supports:\n\n- **Chart extraction** — Chart‑to‑CSV, Chart‑to‑Summary, Chart‑to‑Code \n- **Table extraction** — JSON, HTML, and OTSL \n- **Semantic KVP extraction** — Schema‑guided extraction across diverse document layouts \n- **Image‑to‑text** — Natural‑language descriptions of images \n\nGranite‑Vision-4.1‑4B preserves and extends Granite Vision 4 capabilities while providing more specialized extraction workflows.\n\n---\n\n## 📊 ChartNet Dataset \n\n**ChartNet** is a million‑scale multimodal dataset created to support robust chart understanding tasks: \n➡️ https://huggingface.co/datasets/ibm-granite/ChartNet\n\nIt includes:\n\n- **1.7M** synthetic charts with aligned images, code, tables, summaries, and reasoning \n- **94,643** human‑verified charts \n- **2,000** human‑verified test samples \n- **24 chart types**, across **6 plotting libraries** \n\nChartNet uses a **code‑guided synthe"},{"ref":"P16","kind":"page","title":"ibm-granite/granite-3.3-language-models repository metadata","date":"2026-06-11T04:19:56.465383+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/granite-3.3-language-models","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/granite-3.3-language-models\n\nDescription: Granite 3.3 repository\n\nLicense: Apache-2.0\n\nStars: 23\n\nForks: 5\n\nOpen issues: 2\n\nCreated: 2025-04-15T16:27:46Z\n\nPushed: 2025-06-25T20:42:37Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n\n![](\"figures/granite-3_3-language-models-3x-v1.png\")
\n
\n:books: Paper (comming soon) | :hugs: HuggingFace Collection | \n:speech_balloon: Discussions Page | 📘 IBM Granite Docs & Granite Cookbooks\n
\n\n---\n## Introduction to Granite 3.3 Language Models\nGranite 3.3 language models are lightweight, state-of-the-art, open foundation models that natively support multilinguality, coding, reasoning, and tool usage, including the potential to be run on constrained compute resources. All the models are publicly released under an Apache 2.0 license for both research and commercial use. The models' data curation and training procedure were designed for enterprise usage and customization, with a process that evaluates datasets for governance, risk and compliance (GRC) criteria, in addition to IBM's standard data clearance process and document quality checks.\n\nGranite 3.3 models retain key capabilities of earlier versions, such as a long context support, and features to control the response length and originality through annotations. Additionally, the models introduce fill-in-the-middle (FIM) support for code completion and improve the clarity of model reasoning by separating intermediate thoughts from final answers. Granite 3.3 models are available in two different sizes and are built on a dense architecture.\n\nGranite 3.3 was trained on synthetic data generated from a variety of different open source LLMs, including but not limited to open source models like Mistral and Gemma. Gemma is provided under and subject to th"},{"ref":"P17","kind":"page","title":"ibm-granite/granite-common repository metadata","date":"2026-06-11T03:22:49.187975+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/granite-common","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/granite-common\n\nDescription: Python library that provides enhanced prompt creation and output parsing for IBM Granite models\n\nLanguage: Jupyter Notebook\n\nLicense: Apache-2.0\n\nStars: 10\n\nForks: 8\n\nOpen issues: 11\n\nCreated: 2025-06-27T18:01:34Z\n\nPushed: 2026-05-21T19:20:27Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n# granite-common\n\nPython library that provides enhanced prompt creation and output parsing for IBM\nGranite models.\n\n## Installation\n\nTo install from the main development branch, type:\n\n```\npip install git+https://github.com/ibm-granite/granite-common.git\n```\n\n## Developer setup\n\nFor compatibility with different underlying operating system versions, we recommend using `conda` to create a consistent base Python environment for development and testing.\n\nDetailed instructions:\n\n1. Install [MiniForge](https://github.com/conda-forge/miniforge) or another package that provides the `conda` command-line utility.\n1. Set the environment variable `CONDA_HOME` to point to the root of your `conda` install. If you installed MiniForge in your home directory, this value should be `${HOME}/miniforge3`.\n1. Check out a copy of this repository.\n1. Run the script [`scripts/env.sh`](scripts/env.sh) from the root of your local copy of the repository. The script will create a Conda environment in `./env` and will install the source code of your local copy as an editable Pip package. The script will also install and enable pre-commit hooks with [pre-commit](https://pre-commit.com/).\n1. Before running commands such as `python` or `jupyter` from the command line, activate the Conda environment by typing `conda activate ./env` from the root of your local copy of this repository.\n1. If you are using Visual Studio Code or a similar IDE, configure your IDE to use the environment at `./env`\n\n## Running tests\n\nAfter following the instructions in the previous section, you should be able to run tests on your local machine by typing:\n```\npytest tests\n```\nfrom the root of your local copy of this repository, using the conda environment\ndescribed in the previous section.\n\nThe build automation for this project uses the [`tox`](https://tox.wiki/en) environment manager"},{"ref":"P18","kind":"page","title":"ibm-granite/granite-speech-models repository metadata","date":"2026-06-11T03:22:49.178828+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/granite-speech-models","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/granite-speech-models\n\nLanguage: Jupyter Notebook\n\nStars: 44\n\nForks: 6\n\nOpen issues: 3\n\nCreated: 2025-07-07T21:02:29Z\n\nPushed: 2026-04-28T14:32:26Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n
\n![](\"figures/granite_speech_repo_headbar.png\")
\n
\n:books: Tech Report | :hugs: HuggingFace Collection | \n:trophy: OpenASR leaderboard | \n:wrench: Finetuning Example \n
\n\n# Granite Speech Models \n\n**Model Summary:**\nGranite Speech models are compact and efficient speech-language models, specifically designed for automatic speech recognition (ASR) and automatic speech translation (AST). Granite Speech models use a [two-pass design](https://github.com/ibm-granite/granite-speech-models/blob/main/notebooks/two_pass_spoken_qa.ipynb), unlike integrated models that combine speech and language into a single pass. Initial calls to Granite Speech will transcribe audio files into text. To process the transcribed text using the underlying Granite language model, users must make a second call as each step must be explicitly initiated.\n\nThese models were trained on a collection of public corpora comprising diverse datasets for ASR and AST as well as synthetic datasets tailored to support the speech translation task. granite-speech-3.3-2b/8b models were trained by modality aligning granite-3.3-2b/8b-instruct to speech on publicly available open source corpora containing audio inputs and text targets.\n\n* Compared to revision 3.3.1, revision 3.3.2 supports multilingual speech inputs in English, French, German, Spanish and Portuguese and provides additional accuracy improvements for English ASR.\n* Compared to the initial release, revision 3.3.2 is also trained on additional data and uses a deeper acoustic encoder for improved transcription acc"},{"ref":"P19","kind":"page","title":"ibm-granite/gguf repository metadata","date":"2026-06-11T03:22:49.126284+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/gguf","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/gguf\n\nDescription: CI/CD for IBM model GGUF conversions, quantizations and packagings for partner delivery\n\nLanguage: Jupyter Notebook\n\nLicense: Apache-2.0\n\nStars: 4\n\nForks: 0\n\nOpen issues: 1\n\nCreated: 2025-10-02T20:06:28Z\n\nPushed: 2025-10-17T16:08:16Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n# gguf\n\nThis repository provides an automated CI/CD process to convert, test and deploy IBM Granite models, in safetensor format, from the `ibm-granite` organization to versioned IBM GGUF collections in Hugging Face Hub under the [`ibm-research` organization](https://huggingface.co/collections/ibm-research). This includes:\n\n- [Granite 3.2 Models (GGUF)](https://huggingface.co/collections/ibm-research/granite-32-models-gguf-67bf411f8eb52909dde3532b)\n\n#### Topic index\n\n- [Target IBM models for format conversion](#target-ibm-models-for-format-conversion)\n- [Supported IBM Granite models (GGUF)](#supported-ibm-granite-models-gguf)\n- [Language](#language)\n- [Guardian](#guardian)\n- [Vision](#vision)\n- [Embedding](#embedding-dense)\n- [GGUF Conversion & Quantization](#gguf-conversion--quantization)\n- [GGUF Verification Testing](#gguf-verification-testing)\n- [References](#references)\n- [Releasing GGUF model conversions & quantizations](#releasing-gguf-model-conversions--quantizations)\n\n---\n\n### Target IBM models for format conversion\n\nFormat conversions (i.e., GGUF) and quantizations will only be provided for canonically hosted model repositories hosted in an official IBM Huggingface organization.\n\nCurrently, this includes the following organizations:\n\n- https://huggingface.co/ibm-granite\n- https://huggingface.co/ibm-research\n\nAdditionally, only a select set of IBM models from these orgs. will be converted based upon the following general criteria:\n\n- The IBM GGUF model needs to be referenced by an AI provider service as a \"supported\" model.\n- *For example, a local AI provider service such as [Ollama](https://ollama.com/) or a hosted service such as [Replicate](https://replicate.com/).*\n\n- The GGUF model is referenced by a public blog, tutorial, demo, or other public use case.\n- Specifically, if the model is referenced in an IBM [Granite Snack Cookbook](ht"},{"ref":"P20","kind":"page","title":"ibm-granite/granite-4.0-language-models repository metadata","date":"2026-06-11T03:22:49.033776+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/granite-4.0-language-models","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/granite-4.0-language-models\n\nLicense: Apache-2.0\n\nStars: 213\n\nForks: 24\n\nOpen issues: 10\n\nCreated: 2025-09-17T17:27:15Z\n\nPushed: 2026-03-30T21:18:32Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n
\n![](\"figures/granite-4_0-language-models-3x-v1.png\")
\n
\n:hugs: HuggingFace Collection | \n:speech_balloon: Discussions Page | 📘 IBM Granite Docs\n
\n\n---\n## Overview\nGranite 4.0 language models are lightweight, state-of-the-art open foundation models that natively support multilingual capabilities, a wide range of coding tasks—including fill-in-the-middle (FIM) code completion—retrieval-augmented generation (RAG), tool usage, and structured JSON output.\n\nOur models are developed using a combination of advanced techniques such as structured chat formatting, supervised fine-tuning, reinforcement learning–based model alignment, and model merging. Granite 4.0 features significantly improved *instruction-following* and *tool-calling* capabilities, making it highly effective for enterprise applications and an ideal choice for deployment in environments with constrained compute resources.\n\nAll models are publicly released under the Apache 2.0 license, allowing free use for both research and commercial purposes. The data curation and training processes were specifically designed for enterprise scenarios and customization, incorporating governance, risk, and compliance (GRC) evaluations alongside IBM’s standard data clearance and document quality review procedures.\n\nThe initial release of the Granite 4.0 models included three sizes—micro, tiny, and small—built on dense, dense-hybrid, and mixture-of-experts (MoE) hybrid architectures. Additional model sizes have been added gradually. We provide both base models (checkpoints after pretraining) and instruct models (checkpoints fine-tuned for dialogue, instruction following, helpfulness, and safety).\n\nCore evaluation results for all "},{"ref":"P21","kind":"page","title":"ibm-granite/granite-4.0-nano-language-models repository metadata","date":"2026-06-11T03:22:49.006004+00:00","date_source":null,"source_url":"https://github.com/ibm-granite/granite-4.0-nano-language-models","signal_url":null,"signal_json_url":null,"text":"# ibm-granite/granite-4.0-nano-language-models\n\nLicense: Apache-2.0\n\nStars: 44\n\nForks: 9\n\nOpen issues: 4\n\nCreated: 2025-10-09T21:11:08Z\n\nPushed: 2026-03-30T21:20:13Z\n\nDefault branch: main\n\nFork: no\n\nArchived: no\n\nREADME:\n
\n![](\"figures/granite-4_0-nano-language-models-3x-v1.png\")
\n
\n:hugs: HuggingFace Collection | \n:speech_balloon: Discussions Page | 📘 IBM Granite Docs\n
\n\n---\n## Overview\nGranite 4.0 Nano language models are lightweight, state-of-the-art open foundation models designed for scenarios where efficiency and speed are critical. They can run on resource-constrained devices such as smartphones or IoT hardware, enabling offline and privacy-preserving applications. These models support a wide range of coding tasks—including fill-in-the-middle (FIM) code completion through the use of specialized prefix and suffix tokens, retrieval-augmented generation (RAG), tool usage, and structured JSON output.\n\nAll models are publicly released under the Apache 2.0 license, allowing free use for both research and commercial purposes. The data curation and training processes were specifically designed for enterprise scenarios and customization, incorporating governance, risk, and compliance (GRC) evaluations alongside IBM’s standard data clearance and document quality review procedures.\n\nGranite 4.0 Nano models are available in two sizes, 350M and 1B parameters. This model family supports dense and dense-hybrid architectures. We release both base models (checkpoints after pretraining) and instruct models (checkpoints fine-tuned for dialogue, instruction following, helpfulness, and safety).\n\n## Evaluation Results\nCore evaluation results for all model variants are provided on their respective model cards, and a more comprehensive extended evaluation is available [here](URL).\n\n