IBM Releases Updated Granite 3.1 Open Source Language Model with Significant Performance Improvements

AIbase基地

Published inAI News · 4 min read · Dec 24, 2024

Recently, IBM announced the release of version 3.1 of its open-source language model, Granite, which brings many significant improvements. The new version features a redesigned, denser architecture that can handle up to 128,000 tokens at once. This update means that Granite's ability to process complex texts and tasks has been significantly enhanced.

IBM

The Granite 3.1 model has been trained on datasets covering 12 languages and 116 programming languages, processing a total of 12 trillion tokens. This makes the model's performance in language understanding and generation even more outstanding, better meeting user needs. IBM stated that these new models excel in tasks such as answering questions using external data (RAG), extracting information from unstructured text, and creating document summaries.

Developers can now access these models through the Hugging Face platform, providing powerful support for various application scenarios. Granite was first launched in May 2024, and this update marks IBM's ongoing progress and innovation in the open-source AI field.

IBM's Granite model is not just a technical enhancement but also provides developers and enterprises with more flexible and powerful tools, enabling them to process and analyze data more efficiently. As AI technology continues to evolve, the Granite model will continue to play a vital role in helping various industries achieve digital transformation.

Through this series of improvements, IBM hopes to attract more developers to participate in the open-source community to jointly promote the development of AI technology. The release of Granite 3.1 is not only a technological leap but also a positive push for future research in language models.

Project link: https://huggingface.co/collections/ibm-granite/granite-31-language-models-6751dbbf2f3389bec5c6f02d

Highlights:
🌟 The redesigned Granite 3.1 model can process up to 128,000 tokens.
🌍 The model's training data includes 12 languages and 116 programming languages, totaling 12 trillion tokens processed.
💻 Developers can access these powerful open-source language models through the Hugging Face platform.

Beyond Gemma2! IBM Releases Granite3.1 Model: Supports 128K Context Length

IBM has officially launched its next-generation open-source large language model, Granite3.1, aiming to secure a leading position in the enterprise AI sector. This series of models features an extended context length of 128K, embedding models, built-in hallucination detection capabilities, and significant performance improvements. IBM claims that the Granite8B Instruct model outperforms its open-source competitors of similar scale, including Meta's Llama3.1, Qwen2.5, and Google's Gemma2.

TuSimple Launches Image-to-Video Model 'Ruyi' and Opens Sources Ruyi-Mini-7B

Beijing TuSimple Technology Co., Ltd. officially launched its first 'Image-to-Video' large model - 'Ruyi' on December 17, 2024, and has made the Ruyi-Mini-7B version open source for users to download from the huggingface platform. TuSimple was founded in 2015 and is headquartered in San Diego, California, focusing on the application of AI technology across various industries, including animation gaming and transportation.

Red Hat Launches RHEL AI 1.3: Support for Granite LLM and Intel Gaudi 3

Red Hat recently released the latest version 1.3 of its enterprise-grade Linux AI platform, RHEL AI, which adds support for IBM's Granite large language model (LLM) and previews support for the Intel Gaudi 3 accelerator. This update brings more opportunities for Red Hat's service partners and system integrators to help enterprises explore and integrate AI applications more effectively. Joe Fernandez, Vice President and General Manager of Red Hat's AI business unit, stated that

IBM's Breakthrough Technology: Light Beams Replace Copper Wires, AI Training Efficiency Boosted by 5 Times

In the digital age, computer chips are getting smaller, but computing problems are becoming larger. With the continuous advancement of AI models, energy consumption and computing demands in data centers are growing exponentially. IBM's recent breakthrough in optical technology promises to transform this situation. Traditional data centers face two major technical bottlenecks: the limitations of copper wire transmission and electronic transmission speeds. Currently, most CPUs are largely idle, simply waiting for data packets to be transmitted. This represents a significant waste of computing resources. Image source note: Image generated by AI, image licensed.