<div>
  <p>
  </p>
  <ul>
    <li>
      <p>
        Hugging Face, a platform and community that helps users build, deploy, and train models, has introduced Idefics2, a multimodal model to process text and image inputs and generate text responses.
      </p>
    </li>
    <li>
      <p>
        The model has 8 billion parameters, an open license, and improved OCR (optical character recognition) capabilities. Additionally, the model can answer questions related to images, describe visual content, create stories based on multiple images, extract information from various documents, and execute basic arithmetic operations.
      </p>
    </li>
  </ul>
  <p>
  </p>
</div>


<div>
  <p>
    <a href="https://huggingface.co/blog/idefics2">
      Company blog
    </a>
  </p>
  <p>
  </p>
</div>


Foundation Models

PROMINENT_CATEGORY

<div>
 
 <p>Hugging Face has released Idefics2, a multimodal model designed to process both text and image inputs while generating text responses. The platform provides tools for building, deploying, and training models.</p>
 <p>With 8 billion parameters, an open license, and enhanced optical character recognition (OCR) capabilities, Idefics2 can perform various tasks. These include answering image-related queries, describing visual contents, storytelling based on multiple images, extracting information from different documents, and executing basic arithmetic functions.</p>
 
</div>

Hugging Face releases Idefics2, a multimodal model to process text and image inputs

Exploring the limitless possibilities of AI

Generative AI Applications

<div>

 <p>Cohere has incorporated Command R+ into Hugging Face's HuggingChat, providing enhanced retrieval-augmented generation features and support for multiple languages.</p>
 <p>The advanced Command R+ bolsters real-life business applications with raised performance and is presently accessible on Amazon Sagemaker and Microsoft Azure.</p>
 <p>HuggingChat provides a selection of large language models to users, featuring Command R+ among others, distinguishing it in the realm of AI chatbots and possibly drawing the attention of developers for enterprise utilization.</p>
 
</div>

Hugging Face integrates Cohere’s Command R+ into HuggingChat

The toolkit to build machine learning applications at scale

Machine Learning Infrastructure

<div>
 
 <p>Hugging Face, a machine learning and natural language processing platform, has formed a partnership with Wiz Research, a tech research firm. The collaboration aims to detect and handle possible security threats in Hugging Face's AI cloud services platform.</p>
 <p>With this alliance, Wiz Research is committed to enhancing Hugging Face's security measures, minimizing the potential risks associated with the shared infrastructure used for storing sensitive data.</p>
 
</div>

Hugging Face and Wiz Research partner to enhance security practices

INCUMBENTS_PRESENT

<div>
 
<p>Hugging Face and NVIDIA have joined forces with ServiceNow to launch StarCoder2, an open-source large language model tailored for code generation. The system is proficient in 619 programming languages and can be further adjusted and integrated into company applications to generate application source code, build workflows, and summarise text.</p>
<p>The StarCoder2 family includes a 3-billion-parameter model developed by ServiceNow, a 7-billion-parameter model produced by Hugging Face, and a 15-billion-parameter model made by NVIDIA, which incorporates NVIDIA NeMo and is trained on NVIDIA boosted infrastructure.</p>
<p>The lower-parameter models are cost-effective, as they require less compute power for inference. Moreover, it's reported that the new 3-billion-parameter model matches the performance of the original 15-billion-parameter StarCoder model.</p>

</div>

Hugging Face and NVIDIA partner with ServiceNow to launch StarCoder2 for improved code generation

ANALYST_QUICK_TAKE

<div>
 
 <p>Hugging Face, a modeling platform and community, has introduced a new service named Hugging Chat Assistants, which enables users to craft their customized chatbots. Users can personalize these chatbots with a unique name, avatar, and description.</p>
 <p>The chatbots work using multiple open-source Large Language Models (LLMs) such as Meta's Llama 2 and Mistral's Mixtral, giving users the liberty to select the model that best fits their requirements.</p>
 
</div> <p><b>Analyst QuickTake:<br>
</b> This feature comes as a rival for OpenAI’s <a href="https:<br>
//sp-edge.com/updates/25407"> GPTs </a> while offering users greater flexibility as it comes without any subscription fees. However, it's worth noting that chatbots developed with alternatives like GPT-4 might have additional features beyond those provided by open-source models.&nbsp;</p>

Hugging Face introduces free customizable GenAI chatbots named Hugging Chat Assistants

<div>
 
 <p>Hugging Face, an ML and natural language processing platform, has formed a strategic partnership with Google Cloud. This partnership will empower developers on Hugging Face to utilize the AI-optimized infrastructure of Google Cloud including GPUs and TPUs. Developers can also build GenAI applications and train and deploy Hugging Face models directly on Google Cloud's Vertex AI.</p>
 <p>Furthermore, developers can manage and bill services including inference, endpoints, and AutoTrain through the Google Cloud Marketplace.</p>
 
</div>

Hugging Face and Google Cloud partner to facilitate AI development on cloud infrastructure

Empowering Tomorrow's Innovations with Generative AI Infrastructure

Generative AI Infrastructure

<div>
 <p>Hugging Face, an ML and natural language processing platform, has partnered with Google Cloud. This partnership allows developers to utilize Google Cloud's infrastructure within their Hugging Face projects. Developers can use Google's AI-enhanced resources such as GPUs and TPUs to train and operate open models and to build new GenAI applications. They will also be able to train and deploy Hugging Face models directly from the platform using Google Cloud's Vertex AI. Furthermore, the Google Cloud Marketplace can be utilized to manage the Hugging Face platform, including inference, endpoints and AutoTrain.</p>
</div>

<div>
 <p>Hugging Face, a machine learning and natural language processing platform, has forged a strategic partnership with Google Cloud. This will enable developers on Hugging Face to utilize Google Cloud's infrastructure. Developers can use Google Cloud's AI-optimized facilities like GPUs and TPUs to construct new GenAI applications and to train and serve open models. Plus, they can train and deploy Hugging Face models directly from the Hugging Face platform on Google Cloud's Vertex AI. Google Cloud Marketplace can be used for managing and billing the Hugging Face managed platform, covering inference, endpoints, and AutoTrain.</p>
</div>

<div>
 
 <p>Lasso Security has identified potential security risks in GenAI models and platforms, notably with Hugging Face.</p>
 <p>Hugging Face faced a security threat as Lasso Security discovered 1,681 API tokens at risk of compromise. The tokens, found across GitHub and Hugging Face repositories, posed a serious threat to organizations, including Meta, Microsoft, Google, and others.</p>
 <p>Lasso Security recommended continuous scanning for exposed API tokens and their revocation or notification.&nbsp;</p>
 
</div>

Lasso Security identifies security risks in Hugging Face’s GenAI models and platforms

DISRUPTOR_PRESENT

<div> 
 <ul> 
  <li><p>Dell Technologies is partnering with Hugging Face to facilitate on-premises deployment of Gen AI models, aiming to address enterprise concerns about cloud costs and data security.</p></li> 
  <li><p>The partnership provides a curated library of AI models for selection along with optimized infrastructure choices. The portal includes custom containers and scripts for deploying models on Dell servers and data storage systems.</p></li> 
  <li><p>The collaboration also offers access to datasets, libraries, and tutorials for GenAI model training, streamlining the process for enterprises.</p></li> 
 </ul> 
</div> 

Dell Technologies is partnering with Hugging Face to enable on-premises deployment of Gen AI models. This partnership aims to tackle enterprise concerns regarding the cost of cloud services and data security. The collaboration offers a curated library of AI models, along with optimized infrastructure options. The portal includes custom containers and scripts for deploying models on Dell servers and data storage systems. Additionally, this partnership provides access to datasets, libraries, and tutorials for GenAI model training, simplifying the process for enterprises.

All Updates

Hugging Face releases Idefics2, a multimodal model to process text and image inputs

Hugging Face integrates Cohere’s Command R+ into HuggingChat

Hugging Face and Wiz Research partner to enhance security practices

Hugging Face and NVIDIA partner with ServiceNow to launch StarCoder2 for improved code generation

Hugging Face introduces free customizable GenAI chatbots named Hugging Chat Assistants

Hugging Face and Google Cloud partner to facilitate AI development on cloud infrastructure

Hugging Face and Google Cloud partner to facilitate AI development on cloud infrastructure

Hugging Face and Google Cloud partner to facilitate AI development on cloud infrastructure

Lasso Security identifies security risks in Hugging Face’s GenAI models and platforms

Dell Technologies partners with Hugging Face to provide on-premise deployment of GenAI models

Foundation Models

Hugging Face releases Idefics2, a multimodal model to process text and image inputs

Contact us