share presentation mode

Screen sharing a PowerPoint presentation

There are three methods you can use to screen share a PowerPoint presentation in a Zoom meeting. If you have dual monitors, you can share a slide show while viewing the presenter's notes on another monitor. If you have a single monitor, you can also start the slide show in a window so you have access to other meeting features while sharing your presentation. 

If you have other participants presenting portions of the PowerPoint, you can give them slide control in Zoom, so that they can control the slideshow on their end, without needing to ask you to move the slides forward. Additionally, PowerPoint slides can be shared as a Virtual Background for a more immersive sharing experience. 

This article covers:

Dual monitors with slide show and presenter's views

Single-monitor setup with slide show view in a window, single-monitor setup with slide show in full screen.

Follow these steps if you are using multiple monitors and want to present your PowerPoint in one monitor, while viewing the presenter's notes in another monitor.

• Open the PowerPoint file you want to present.
• Start or join a Zoom meeting.

• Select your primary monitor then click Share . If you are not sure which monitor is your primary, select the one that PowerPoint opens in.

• Switch back to Powerpoint and click the Slide Show tab. 

Follow these steps if you have a single monitor and want to share your PowerPoint presentation in slide show view, but have it contained in a window rather than in full screen. This is useful if you need to access meeting features, such as in-meeting chat or managing participants, while sharing your PowerPoint presentation.

• Click the Slide Show tab and then select Set Up Slide Show .
• Under Show type , select Browsed by an individual (window) and then click OK .

• In Zoom, start or join a meeting .
• Select the PowerPoint window and then click Share .

Note : Be sure you select the PowerPoint window, not the entire screen. Sharing the PowerPoint window only will allow you to use other features without interrupting the view of the presentation. 

• Select your monitor then click Share . 

AppsThatDeliver

How-Tos, B2B Software Ratings & Reviews

How to Share Google Slides in Presentation Mode

Would you like to learn how to share Google Slides in presentation mode?

When you copy the file link and share it with others, the file typically opens in edit mode.

However, what if you want to share it in presentation mode?

Fortunately, it is possible to do so in Google Slides!

In this article, you will learn how to share Google Slides in presentation mode, step-by-step.

Also read: (opens in new tab) How to Track Changes in Google Slides How to Time Google Slides for 20 Seconds Google Slides Semi Transparent Shape Google Slides Animate Bullets How to Lock an Image in Google Slides

How to Share Google Slides in Presentation Mode?

Step 1: click on the “share” button.

To get started, open the Google Slides file.

Once opened, you’ll find the “Share” button located at the top right corner of the screen.

Click on the “Share” button, and then proceed to the next step.

Step 2: Under “General access” select “Anyone with the link”

A pop-up window with sharing options will appear.

By default, the “Restricted” option will be selected under the “General access” settings.

Click on this option, and then choose “Anyone with the link” from the drop-down menu.

Ensure that the role is set to “Viewer”.

Once done, proceed to the next step.

Step 3: Click “Copy link”

On the pop-up, you will see the “Copy link” button in the bottom left corner.

Click on this button to copy the presentation link to your clipboard.

However, do not share this link yet.

We need to make some changes to the link so that the slides open in presentation mode instead of edit mode.

Step 4: Paste the link in a text editor

After clicking the “Copy link” button, the link will be copied to your clipboard.

Paste the link into your favorite text editor so that you can edit it.

Alternatively, you can paste the link into the URL bar of your browser, but do not press the enter key on your keyboard; simply paste the link and proceed to the next step.

Step 5: Replace “edit?usp=sharing” with “present?usp=sharing”

The pasted link should look something like this:

Now, focus on the last part of the link.

You’ll notice the text edit?usp=sharing . Replace that text with present?usp=sharing

After making the change, your final link should look something like this:

Step 6: Share the link

Before sharing the link, make sure to open it in your browser.

Check if the slides load properly, especially in presentation mode.

If the slides open correctly, you are all set to share the link.

Yes, you can share a Google Slide in presentation mode. To do so, click the “Share” button, select “Anyone with the link” and “Viewer” role, copy the link, replace “edit” with “present” in the URL, and share.

To share a Google Slide presentation in presentation mode, click the “Share” button, select “Anyone with the link” and “Viewer” role, copy the link, replace “edit” with “present” in the URL and share.

Sharing your Google Slides in presentation mode is a simple process that can make all the difference in how your audience experiences your presentation.

By following these steps, you can share your Google Slides in presentation mode:

• Click on the “Share” button
• Under “General access,” select “Anyone with the link”
• Click “Copy link”
• Paste the link into a text editor
• Replace “edit?usp=sharing” with “present?usp=sharing”
• Share the link

Remember, always check if the slides load properly before sharing the link.

Author: Shubham Calmblay

Shubham Calmblay, founder of appsthatdeliver.com, has a decade of experience with various Google products. He has authored 1,000+ guides for ATD, published on prestigious tech blogs. His work has garnered recognition from Protocol.com, Leadsbridge.com, MadMobile.com, and numerous other leading publications and corporations.

All Posts by Shubham Calmblay

How to share your Google Slides presentation

• Written by: Emma Trantham
• Categories: Google Slides
• Comments: 4

Wondering how to share a Google Slides presentation? Good news! It’s a pretty simple process. There are multiple ways of sharing a Google Slides deck. The right method for you will depend on how you’re going to deliver the slides, the required file type, and your audience. Do you want to share so that a colleague can proof-read and edit, or so that an online audience has the virtual equivalent of a handout?

As Google Slides exists completely online, sharing your presentation with others isn’t limited to an email attachment. This post will show you how to share Google Slides presentations in five ways, there’s definitely something here to meet your needs!

Share with people and groups

Publish to the web

Download as different file type

Email as attachment

Before we get to that, you need to know where to find the Share function in Google Slides.

• Open Google Slides and click File , in the top bar
• Click Share at the top of the drop-down menu

• Or, click Share at the top right-hand side of your window.

1. How to share Google Slides presentations with people and groups

Using Share with people and groups , you can share your Google Slides presentation with multiple people at once. This is helpful if your teammates need to collaborate in the same deck or if you require feedback from a colleague or client. Share with people and groups allows you to change the editing permissions per contact. You can allow each contact to simply view the document, or add comments, or edit it themselves. The level of permission you choose can be different for each contact you share with, giving you a lot of flexibility.

• Click Share to bring up the Share with people and groups pane.
• Next to Add people and groups , start typing the name or email address of the contact you want to share the deck with. A list of recommendations will appear as you type. Once you’ve selected the right contact, their name will appear at the top of the pane. If the person you want to share with is not already in your contact list, you need to type in their full email address.
• Click the small cross next to a contact name to deselect the contact.

When you share using Share with people and groups , Google will ask the user to log in to their Google account to access your shared file. If the user does not have a Google account, they will be unable to access your deck. Just head to the next section to find out how to grant access to non-Google users.

Now you’ve chosen your contact, you can alter their permissions to change how they are able to interact with the deck. There are three options – Viewer , Commenter , and Editor . Simply click on the option currently visible and a drop-down menu will appear allowing you to choose whether your contact can… well… view, comment on or edit the deck. Pretty self-explanatory, right?

But which permission should you give to who and when?

• Viewer: This will prevent contacts from editing slides. If your deck is completely finished and approved, set the permissions to Viewer to stop contacts accidentally editing – or even deleting – slides from your finished presentation!
• Commenter: The best option for gathering feedback. If your contact has limited knowledge of Google Slides, direct them towards our handy Ultimate Guide to Google Slides post so that they can brush up! In the meantime, share your deck using the Commenter function. This way, contacts can comment without being able to rearrange slides, or change text or animations.
• Editor: Perfect for collaborative projects. Editor allows you to share an editable version of your deck. Everyone with this permission can work on the same slides at the same time. This means you don’t end up with multiple versions of the same deck.

For more tips on how to improve collaboration in Google Slides, check out these 6 ways to get the most out of online collaboration with Google Slides .

Once you’ve selected one of these permission options, you’re free to hit Send . If you tick Notify people , your contact will receive an email letting them know that they have permission to access or interact with your deck.

In the Share with people and groups pane, the owner of the Google Slides presentation appears at the top. Beneath that, you can see who has access to the deck and the status of their permissions. If you want to change someone’s permissions, press the yellow Share button at any time and edit permissions using the same drop-down menu.

To stop sharing with a specific person:

• Click Share and scroll to the contact you want to remove
• Click the drop-down arrow next to their name
• Select Remove
• Click Save .

Notice the Give temporary access and Make owner options. If you choose to Give temporary access , you can edit the number of days that your contact has access to your deck. All you have to do is click the number of days next to Access expires to change the expiration date. This can be a great option if you’re working with people outside of your organization, such as clients or agencies. For example, you can schedule the expiration date for the end of a project, once the feedback has been received and the work completed.

Make owner allows you to pass ownership of the deck to another user. After you’ve created a presentation you might need to hand it over, to the presenter for example. You can use this option to give someone else complete control. Use it with caution, however. Once you’ve handed over ownership, you cannot retrieve it – unless the new owner hands it back or chooses to grant you access permissions.

Top tip: If you’re the owner of the deck (if you have created it yourself or have been given ownership by the original creator), there are further sharing options. When you click Share , you’ll notice a cog appears in the top right-hand corner pop up. This allows you to choose whether your Editors can change permissions and share or if Viewers and commenters can see the option to download, print, and copy. All you have to do is tick or untick the box. Easy!

2. How to share a Google Slides presentation via a link

In this window you can also get a link.

• Click the blue Copy link button.
• Paste the link anywhere, from an email draft to a Microsoft Teams chat

This link will only work for people who have been given permission to view, edit or comment on the deck. To change this:

• Click Share
• In the Get link section, press Change .
• Choose whether your deck is Restricted or if Anyone with the link can view. Note: when you change a link’s setting to Restricted , the only people who can still see the deck are those you have directly shared with in the Share with people and groups section.

Though the Anyone with link function can be useful, it’s not the most secure option. Remember: the users that you originally shared the link with can continue to share the link with their friends, colleagues, and contacts too, meaning your presentation can be shared endlessly and with anyone! If your deck contains confidential content, you should not use this method.

3. How to publish a Google Slides presentation to the web

If you don’t want your viewers to edit your deck, then Share with groups and people may not be the option for you. Instead, choose to Publish to the web . With this option, your deck becomes non-editable, it will simply play as if your deck was a video; auto-advancing through the slides. This means that the viewer won’t have control over which slides they view or the order they view them in. The content of your slides will be visible to anyone and everyone when publishing to the web, so be careful about choosing this option if your slides contain confidential information.

• Click  File
• Select Publish to the web
• Choose Link or Embed. You can share the URL with your contacts, or you can embed the link on a website or blog
• Choose how your slides playback using the drop-down menu beneath Auto-advance slides
• Click Publish
• Copy the URL or HTML and share your deck

4. How to download a Google Slides deck as different file type

Another method of sharing (or saving) your slides involves downloading the deck as a different type of file.

• Under File select Download .
• Select the format you need (Microsoft PowerPoint (.pptx), PDF Document (.pdf), or JPEG Image (.jpeg)), and the download will begin.
• Open your file

While there are many formats available, we don’t recommend all of them. For example, when downloading as an Image, whether PNG or JPEG, the quality of your slides can be lost, the presentation will be static, and you can only download one slide at a time. If you’re printing or emailing your deck, downloading as a PDF is the best option to get a static version of your slides.

Since PowerPoint is the closest option to Slides, this format retains the presentability of the deck. However, there’s a chance that your content will corrupt or break as not all features are transferable between file formats. Check for any changes and make tweaks – like removing animations – before sending your presentation to your contacts. To find out more about the differences between PowerPoint and Slides, check out 9 Google Slides features we wish PowerPoint had or if you want to try different platforms to PowerPoint, we recommend The ULTIMATE Guide to PowerPoint Alternatives .

5. How to email your Google Slides presentations as an attachment

  If all these sharing options still aren’t quite cutting it, there’s one more! To share your slides via email:

• Select Email as attachment
• Type the name or email address of your contact
• Pick the file type your deck
• Add a message, then hit Send

That’s the last of our five tips on how to share a Google Slides presentation. Now you know pretty much everything there is to know about sharing your slides and, if you want to make those slides even more awesome and effective, you have got to check out these 5 Google Slides video tutorials to level up your skills .

Emma Trantham

Principal consultant, related articles, how to present in google slides with present mode toolbar.

• Google Slides / Presentation skills

It takes time to create engaging, visual slides, so why undermine all your effort at the crunch point? Delivering a presentation properly really isn’t that tricky! With Google Slides’ Present mode toolbar, it’s super easy to present in Google Slides navigating your presentation effortlessly.

How to use Presenter view in Google Slides

• Comments: 5

For seasoned presenters and newbies alike, the move to online presenting comes with the joint complications of an online audience and unfamiliar tech. Talk about spinning plates! This blog post covers how to use Presenter view in Google Slides, so that you can be more professional and feel more confident.

The best free Google Slides templates

• Google Slides

Google Slides templates are a great starting point to improve the look and feel of your presentation. They’re fab as they’re accessible and low cost, but it’s important to remember that even well-designed presentations can be ineffective if the content is text heavy.

What is the name of piece that I attach to my chromebook that allows me to present Google slides

Hi Milton. I’m not sure what you mean – could you please clarify?

Hello, My PowerPoint contains an audio file. I uploaded it to Google Sites (new) and there’s no audio. So I uploaded my PowerPoint to Google Slides and inserted that into my Google portfolio Site hoping that would fix the problem. It didn’t. Can you tell me how to get the audio file to upload to my portfolio site, along with my slides? Thank you! Pam

Hi, I am looking for a way to have Google slides running on a laptop or tablet and instead of the presentation going to a screen, to go live i.e. as it would on a projecter in a URL that I can tell people about, so they can use it with their mobile phones, tablets etc. The sort of application would be for Carol singing, each slide would have no more than two lines to allow for large type and as it goes through the verses it shows on singers devices, one slide at a time. This of course would be outside, i.e. people would at best use mobile data, or if it would work for a larger number by wifi tethering (wifi provided by 1phone/tablet. Thank you, Sam

Join the BrightCarbon mailing list for monthly invites and resources

I wanted to make sure I send you both a HUGE thank you for making this story come to life and creating amazing graphics to help. We really appreciate BrightCarbon for stepping up our presentation game massively! Sarah Walker Softchoice

How-To Geek

How to share your powerpoint presentation.

There are other ways to share your presentation besides email. Upload it to the cloud and get a shareable link!

Quick Links

Sharing your powerpoint presentation.

Aside from the traditional method of sending a PowerPoint presentation to others as an email attachment, you can also upload and share your presentation from the cloud. All you need to get started is a OneDrive account. Here’s how.

You can save a PowerPoint presentation to the cloud fairly easily. In fact, it takes only a few simple clicks. For this to work, though, you’ll need a OneDrive account. If you use Office 365, then you already have one. If not, all you need to do is create a Microsoft account  and then sign in to OneDrive. This article will assume you already have a OneDrive account.

Related: How to Share Things from OneDrive

First, open the PowerPoint presentation that you’d like to share. At the top-right corner of the window, you’ll see a “Share” button. Go ahead and select it.

Once selected, the “Share” window will appear. You have a few different options here. In the “Attach a copy instead” group, you can choose to send your presentation as a PowerPoint or PDF attachment. Selecting either of these options will open your computer’s default mail client.

What we’re interested in, though, is sharing to OneDrive. To do this, select your OneDrive account under “Share.”

If you haven’t already named your presentation, you’ll be prompted to do so. After you give it a name, click  “OK.”

Your presentation will now be uploaded to the cloud and the “Share” pane will appear on the right-hand side. This is where you can invite people to edit (or read) the document. In the address bar, enter the email of the recipient. Alternatively, select the icon to the right of the address bar and select a recipient from your Outlook address book.

After you’ve entered the recipient’s email, you can then assign a permission level. Once you’ve given read/write or read-only permission, you can then add an optional message. When you’re ready, click “Share.”

The recipient will receive an email providing access to the presentation.

Note: If you’re the recipient, be sure to check your spam folder! During our testing, we found the invitation in spam.

An alternative method for inviting people is to get a sharing link. At the bottom of the “Share” pane, select “Get a sharing link.”

Next, choose which type of link you’d like to provide. You can choose between a read/write link or read-only link.

A link will then be generated. Select “Copy” to copy the link to your clipboard.

Anyone you share this link with will then have access to the presentation. This will allow multiple people to collaborate on your presentation in real-time !

Present from PowerPoint Live in Microsoft Teams

PowerPoint Live in Teams gives both the presenter and audience an inclusive and engaging experience, combining the best parts of presenting in PowerPoint with the connection and collaboration of a Microsoft Teams meeting.

When you’re the presenter, you have a unique view that lets you control your presentation while staying engaged with your audience, seeing people’s video, raised hands, reactions, and chat as needed.

And if you’re an audience member, you can interact with the presentation and personalize your viewing experience with captions, high contrast slides, and slides translated into your native language.

Here’s how it works:  

Tip:  Are you an audience member? Jump down to learn more about how you can interact during the presentation.

Presenter view

Present your slides

If you're in PowerPoint for the web, select Present > Present in Teams .

Your slides will appear in the Teams meeting, with your Notes next to them.

Navigate through the slides

Use the navigation arrows to go forward and backward.

Use the thumbnail strip to jump ahead or backwards.

Select Go to slide to see a grid view of all slides in the presentation. Select one to jump to it.

Stay connected to the audience

One of the benefits of using PowerPoint Live to present instead of sharing your screen is that you have quick access to all your meeting tools you need to engage with the audience and to read the room in one view. This is especially true if you’re presenting from a single screen.

Turn Chat on or off to view what your audience is saying.

See audience reactions and raised hands in real-time.

Change the Layout of your presentation and choose how your live camera feed appears in your presentation, like Standout or Cameo . It helps the audience read your non-verbal cues and keeps them engaged.

Use the Laser pointer , Pen , Highlighter , or Eraser to clearly reference items on your slides.

Magnifying and panning

As you present, you can zoom in and out and pan around your slides to call attention to specific points and keep your presentation dynamic.

To zoom in or out on a slide, do any one of the following: 

Hover over the slideshow and pinch or stretch on trackpad.

Pinch or use the stretch touch gesture (on a touch-enabled device).

Press the + or – keys.

Hover over slide, hold down Ctrl key and scroll with mouse wheel.

In the More Actions menu, click the + or – buttons.

To pan around your slide, do any one of the following:

Press the arrow keys.

Click and drag using a mouse.

Click and drag on a trackpad.

Use one finger to touch and drag (on touch-enabled device).

When done zooming and panning, press  Esc to reset your screen. 

Audience view

As an audience member, you’re able to personalize your experience without affecting anyone else. Try these options to find what works best for you:

Note:  If presenters don't want people to be able to independently navigate through a PowerPoint file they are sharing, use the  Private view  toggle to turn it off.

Click any hyperlink on slides to get more context right away.

Interact with videos on slides to adjust the volume or jump to a timestamp and consume it at your own pace.

Use a screen reader to get full access to the slide content.

Switch to a high contrast view to make the slides easier to view if you have low vision. Select More options > View slides in high contrast .

Your viewing experience will be at a higher fidelity, letting you see crisp text and smooth animations. PowerPoint Live also requires significantly less network bandwidth than typical sharing, making it the b est option when network connectivity is a problem.

Independent magnifying and panning

You can zoom in and pan on a presentation slide without affecting what others see. Use your mouse, trackpad, keyboard, touch, or the Magnify Slide option as applicable. 

When done zooming and panning, press  Esc to reset your screen.   

Important: 

PowerPoint Live is not supported in Teams live events, CVI devices, and VTC devices.

If you're using Teams on the web, you’ll need Microsoft Edge 18 or later, or Google Chrome 65 or later, to see the presenter view.

Presenter view is hidden by default for small screen devices but can be turned on by selecting More options below the current slide and then Show presenter view (or by selecting the sharing window and then pressing Ctrl+Shift+x).

Meetings recordings won’t capture any videos, animations, or annotation marks in the PowerPoint Live session.

When you share from Teams, the PowerPoint Live section lists the most recent files you've opened or edited in your team SharePoint site or your OneDrive. If you select one of these files to present, all meeting participants will be able to view the slides during the meeting. Their access permissions to the file outside of the meeting won't change.

If you select Browse and choose to present a PowerPoint file that hasn't been uploaded to Teams before, it will get uploaded as part of the meeting. If you're presenting in a channel meeting, the file is uploaded to the Files tab in the channel, where all team members will have access to it. If you're presenting in a private meeting, the file is uploaded to your OneDrive, where only the meeting participants will be able to access it.

Need more help?

Want more options.

Explore subscription benefits, browse training courses, learn how to secure your device, and more.

Microsoft 365 subscription benefits

Microsoft 365 training

Microsoft security

Accessibility center

Communities help you ask and answer questions, give feedback, and hear from experts with rich knowledge.

Ask the Microsoft Community

Microsoft Tech Community

Windows Insiders

Microsoft 365 Insiders

Was this information helpful?

Thank you for your feedback.

• Help Center
• Privacy Policy
• Terms of Service
• Submit feedback
• Announcements

• SUGGESTED TOPICS
• The Magazine
• Newsletters
• Managing Yourself
• Managing Teams
• Work-life Balance
• The Big Idea
• Data & Visuals
• Reading Lists
• Case Selections
• HBR Learning
• Topic Feeds
• Account Settings
• Email Preferences

How to Make a “Good” Presentation “Great”

• Guy Kawasaki

Remember: Less is more.

A strong presentation is so much more than information pasted onto a series of slides with fancy backgrounds. Whether you’re pitching an idea, reporting market research, or sharing something else, a great presentation can give you a competitive advantage, and be a powerful tool when aiming to persuade, educate, or inspire others. Here are some unique elements that make a presentation stand out.

• Fonts: Sans Serif fonts such as Helvetica or Arial are preferred for their clean lines, which make them easy to digest at various sizes and distances. Limit the number of font styles to two: one for headings and another for body text, to avoid visual confusion or distractions.
• Colors: Colors can evoke emotions and highlight critical points, but their overuse can lead to a cluttered and confusing presentation. A limited palette of two to three main colors, complemented by a simple background, can help you draw attention to key elements without overwhelming the audience.
• Pictures: Pictures can communicate complex ideas quickly and memorably but choosing the right images is key. Images or pictures should be big (perhaps 20-25% of the page), bold, and have a clear purpose that complements the slide’s text.
• Layout: Don’t overcrowd your slides with too much information. When in doubt, adhere to the principle of simplicity, and aim for a clean and uncluttered layout with plenty of white space around text and images. Think phrases and bullets, not sentences.

As an intern or early career professional, chances are that you’ll be tasked with making or giving a presentation in the near future. Whether you’re pitching an idea, reporting market research, or sharing something else, a great presentation can give you a competitive advantage, and be a powerful tool when aiming to persuade, educate, or inspire others.

• Guy Kawasaki is the chief evangelist at Canva and was the former chief evangelist at Apple. Guy is the author of 16 books including Think Remarkable : 9 Paths to Transform Your Life and Make a Difference.

Partner Center

Our approach

• Responsibility
• Infrastructure
• Try Meta AI

RECOMMENDED READS

• 5 Steps to Getting Started with Llama 2
• The Llama Ecosystem: Past, Present, and Future
• Introducing Code Llama, a state-of-the-art large language model for coding
• Meta and Microsoft Introduce the Next Generation of Llama
• Today, we’re introducing Meta Llama 3, the next generation of our state-of-the-art open source large language model.
• Llama 3 models will soon be available on AWS, Databricks, Google Cloud, Hugging Face, Kaggle, IBM WatsonX, Microsoft Azure, NVIDIA NIM, and Snowflake, and with support from hardware platforms offered by AMD, AWS, Dell, Intel, NVIDIA, and Qualcomm.
• We’re dedicated to developing Llama 3 in a responsible way, and we’re offering various resources to help others use it responsibly as well. This includes introducing new trust and safety tools with Llama Guard 2, Code Shield, and CyberSec Eval 2.
• In the coming months, we expect to introduce new capabilities, longer context windows, additional model sizes, and enhanced performance, and we’ll share the Llama 3 research paper.
• Meta AI, built with Llama 3 technology, is now one of the world’s leading AI assistants that can boost your intelligence and lighten your load—helping you learn, get things done, create content, and connect to make the most out of every moment. You can try Meta AI here .

Today, we’re excited to share the first two models of the next generation of Llama, Meta Llama 3, available for broad use. This release features pretrained and instruction-fine-tuned language models with 8B and 70B parameters that can support a broad range of use cases. This next generation of Llama demonstrates state-of-the-art performance on a wide range of industry benchmarks and offers new capabilities, including improved reasoning. We believe these are the best open source models of their class, period. In support of our longstanding open approach, we’re putting Llama 3 in the hands of the community. We want to kickstart the next wave of innovation in AI across the stack—from applications to developer tools to evals to inference optimizations and more. We can’t wait to see what you build and look forward to your feedback.

Our goals for Llama 3

With Llama 3, we set out to build the best open models that are on par with the best proprietary models available today. We wanted to address developer feedback to increase the overall helpfulness of Llama 3 and are doing so while continuing to play a leading role on responsible use and deployment of LLMs. We are embracing the open source ethos of releasing early and often to enable the community to get access to these models while they are still in development. The text-based models we are releasing today are the first in the Llama 3 collection of models. Our goal in the near future is to make Llama 3 multilingual and multimodal, have longer context, and continue to improve overall performance across core LLM capabilities such as reasoning and coding.

State-of-the-art performance

Our new 8B and 70B parameter Llama 3 models are a major leap over Llama 2 and establish a new state-of-the-art for LLM models at those scales. Thanks to improvements in pretraining and post-training, our pretrained and instruction-fine-tuned models are the best models existing today at the 8B and 70B parameter scale. Improvements in our post-training procedures substantially reduced false refusal rates, improved alignment, and increased diversity in model responses. We also saw greatly improved capabilities like reasoning, code generation, and instruction following making Llama 3 more steerable.

*Please see evaluation details for setting and parameters with which these evaluations are calculated.

In the development of Llama 3, we looked at model performance on standard benchmarks and also sought to optimize for performance for real-world scenarios. To this end, we developed a new high-quality human evaluation set. This evaluation set contains 1,800 prompts that cover 12 key use cases: asking for advice, brainstorming, classification, closed question answering, coding, creative writing, extraction, inhabiting a character/persona, open question answering, reasoning, rewriting, and summarization. To prevent accidental overfitting of our models on this evaluation set, even our own modeling teams do not have access to it. The chart below shows aggregated results of our human evaluations across of these categories and prompts against Claude Sonnet, Mistral Medium, and GPT-3.5.

Preference rankings by human annotators based on this evaluation set highlight the strong performance of our 70B instruction-following model compared to competing models of comparable size in real-world scenarios.

Our pretrained model also establishes a new state-of-the-art for LLM models at those scales.

To develop a great language model, we believe it’s important to innovate, scale, and optimize for simplicity. We adopted this design philosophy throughout the Llama 3 project with a focus on four key ingredients: the model architecture, the pretraining data, scaling up pretraining, and instruction fine-tuning.

Model architecture

In line with our design philosophy, we opted for a relatively standard decoder-only transformer architecture in Llama 3. Compared to Llama 2, we made several key improvements. Llama 3 uses a tokenizer with a vocabulary of 128K tokens that encodes language much more efficiently, which leads to substantially improved model performance. To improve the inference efficiency of Llama 3 models, we’ve adopted grouped query attention (GQA) across both the 8B and 70B sizes. We trained the models on sequences of 8,192 tokens, using a mask to ensure self-attention does not cross document boundaries.

Training data

To train the best language model, the curation of a large, high-quality training dataset is paramount. In line with our design principles, we invested heavily in pretraining data. Llama 3 is pretrained on over 15T tokens that were all collected from publicly available sources. Our training dataset is seven times larger than that used for Llama 2, and it includes four times more code. To prepare for upcoming multilingual use cases, over 5% of the Llama 3 pretraining dataset consists of high-quality non-English data that covers over 30 languages. However, we do not expect the same level of performance in these languages as in English.

To ensure Llama 3 is trained on data of the highest quality, we developed a series of data-filtering pipelines. These pipelines include using heuristic filters, NSFW filters, semantic deduplication approaches, and text classifiers to predict data quality. We found that previous generations of Llama are surprisingly good at identifying high-quality data, hence we used Llama 2 to generate the training data for the text-quality classifiers that are powering Llama 3.

We also performed extensive experiments to evaluate the best ways of mixing data from different sources in our final pretraining dataset. These experiments enabled us to select a data mix that ensures that Llama 3 performs well across use cases including trivia questions, STEM, coding, historical knowledge, etc.

Scaling up pretraining

To effectively leverage our pretraining data in Llama 3 models, we put substantial effort into scaling up pretraining. Specifically, we have developed a series of detailed scaling laws for downstream benchmark evaluations. These scaling laws enable us to select an optimal data mix and to make informed decisions on how to best use our training compute. Importantly, scaling laws allow us to predict the performance of our largest models on key tasks (for example, code generation as evaluated on the HumanEval benchmark—see above) before we actually train the models. This helps us ensure strong performance of our final models across a variety of use cases and capabilities.

We made several new observations on scaling behavior during the development of Llama 3. For example, while the Chinchilla-optimal amount of training compute for an 8B parameter model corresponds to ~200B tokens, we found that model performance continues to improve even after the model is trained on two orders of magnitude more data. Both our 8B and 70B parameter models continued to improve log-linearly after we trained them on up to 15T tokens. Larger models can match the performance of these smaller models with less training compute, but smaller models are generally preferred because they are much more efficient during inference.

To train our largest Llama 3 models, we combined three types of parallelization: data parallelization, model parallelization, and pipeline parallelization. Our most efficient implementation achieves a compute utilization of over 400 TFLOPS per GPU when trained on 16K GPUs simultaneously. We performed training runs on two custom-built 24K GPU clusters . To maximize GPU uptime, we developed an advanced new training stack that automates error detection, handling, and maintenance. We also greatly improved our hardware reliability and detection mechanisms for silent data corruption, and we developed new scalable storage systems that reduce overheads of checkpointing and rollback. Those improvements resulted in an overall effective training time of more than 95%. Combined, these improvements increased the efficiency of Llama 3 training by ~three times compared to Llama 2.

Instruction fine-tuning

To fully unlock the potential of our pretrained models in chat use cases, we innovated on our approach to instruction-tuning as well. Our approach to post-training is a combination of supervised fine-tuning (SFT), rejection sampling, proximal policy optimization (PPO), and direct preference optimization (DPO). The quality of the prompts that are used in SFT and the preference rankings that are used in PPO and DPO has an outsized influence on the performance of aligned models. Some of our biggest improvements in model quality came from carefully curating this data and performing multiple rounds of quality assurance on annotations provided by human annotators.

Learning from preference rankings via PPO and DPO also greatly improved the performance of Llama 3 on reasoning and coding tasks. We found that if you ask a model a reasoning question that it struggles to answer, the model will sometimes produce the right reasoning trace: The model knows how to produce the right answer, but it does not know how to select it. Training on preference rankings enables the model to learn how to select it.

Building with Llama 3

Our vision is to enable developers to customize Llama 3 to support relevant use cases and to make it easier to adopt best practices and improve the open ecosystem. With this release, we’re providing new trust and safety tools including updated components with both Llama Guard 2 and Cybersec Eval 2, and the introduction of Code Shield—an inference time guardrail for filtering insecure code produced by LLMs.

We’ve also co-developed Llama 3 with torchtune , the new PyTorch-native library for easily authoring, fine-tuning, and experimenting with LLMs. torchtune provides memory efficient and hackable training recipes written entirely in PyTorch. The library is integrated with popular platforms such as Hugging Face, Weights & Biases, and EleutherAI and even supports Executorch for enabling efficient inference to be run on a wide variety of mobile and edge devices. For everything from prompt engineering to using Llama 3 with LangChain we have a comprehensive getting started guide and takes you from downloading Llama 3 all the way to deployment at scale within your generative AI application.

A system-level approach to responsibility

We have designed Llama 3 models to be maximally helpful while ensuring an industry leading approach to responsibly deploying them. To achieve this, we have adopted a new, system-level approach to the responsible development and deployment of Llama. We envision Llama models as part of a broader system that puts the developer in the driver’s seat. Llama models will serve as a foundational piece of a system that developers design with their unique end goals in mind.

Instruction fine-tuning also plays a major role in ensuring the safety of our models. Our instruction-fine-tuned models have been red-teamed (tested) for safety through internal and external efforts. ​​Our red teaming approach leverages human experts and automation methods to generate adversarial prompts that try to elicit problematic responses. For instance, we apply comprehensive testing to assess risks of misuse related to Chemical, Biological, Cyber Security, and other risk areas. All of these efforts are iterative and used to inform safety fine-tuning of the models being released. You can read more about our efforts in the model card .

Llama Guard models are meant to be a foundation for prompt and response safety and can easily be fine-tuned to create a new taxonomy depending on application needs. As a starting point, the new Llama Guard 2 uses the recently announced MLCommons taxonomy, in an effort to support the emergence of industry standards in this important area. Additionally, CyberSecEval 2 expands on its predecessor by adding measures of an LLM’s propensity to allow for abuse of its code interpreter, offensive cybersecurity capabilities, and susceptibility to prompt injection attacks (learn more in our technical paper ). Finally, we’re introducing Code Shield which adds support for inference-time filtering of insecure code produced by LLMs. This offers mitigation of risks around insecure code suggestions, code interpreter abuse prevention, and secure command execution.

With the speed at which the generative AI space is moving, we believe an open approach is an important way to bring the ecosystem together and mitigate these potential harms. As part of that, we’re updating our Responsible Use Guide (RUG) that provides a comprehensive guide to responsible development with LLMs. As we outlined in the RUG, we recommend that all inputs and outputs be checked and filtered in accordance with content guidelines appropriate to the application. Additionally, many cloud service providers offer content moderation APIs and other tools for responsible deployment, and we encourage developers to also consider using these options.

Deploying Llama 3 at scale

Llama 3 will soon be available on all major platforms including cloud providers, model API providers, and much more. Llama 3 will be everywhere .

Our benchmarks show the tokenizer offers improved token efficiency, yielding up to 15% fewer tokens compared to Llama 2. Also, Group Query Attention (GQA) now has been added to Llama 3 8B as well. As a result, we observed that despite the model having 1B more parameters compared to Llama 2 7B, the improved tokenizer efficiency and GQA contribute to maintaining the inference efficiency on par with Llama 2 7B.

For examples of how to leverage all of these capabilities, check out Llama Recipes which contains all of our open source code that can be leveraged for everything from fine-tuning to deployment to model evaluation.

What’s next for Llama 3?

The Llama 3 8B and 70B models mark the beginning of what we plan to release for Llama 3. And there’s a lot more to come.

Our largest models are over 400B parameters and, while these models are still training, our team is excited about how they’re trending. Over the coming months, we’ll release multiple models with new capabilities including multimodality, the ability to converse in multiple languages, a much longer context window, and stronger overall capabilities. We will also publish a detailed research paper once we are done training Llama 3.

To give you a sneak preview for where these models are today as they continue training, we thought we could share some snapshots of how our largest LLM model is trending. Please note that this data is based on an early checkpoint of Llama 3 that is still training and these capabilities are not supported as part of the models released today.

We’re committed to the continued growth and development of an open AI ecosystem for releasing our models responsibly. We have long believed that openness leads to better, safer products, faster innovation, and a healthier overall market. This is good for Meta, and it is good for society. We’re taking a community-first approach with Llama 3, and starting today, these models are available on the leading cloud, hosting, and hardware platforms with many more to come.

Try Meta Llama 3 today

We’ve integrated our latest models into Meta AI, which we believe is the world’s leading AI assistant. It’s now built with Llama 3 technology and it’s available in more countries across our apps.

You can use Meta AI on Facebook, Instagram, WhatsApp, Messenger, and the web to get things done, learn, create, and connect with the things that matter to you. You can read more about the Meta AI experience here .

Visit the Llama 3 website to download the models and reference the Getting Started Guide for the latest list of all available platforms.

You’ll also soon be able to test multimodal Meta AI on our Ray-Ban Meta smart glasses.

As always, we look forward to seeing all the amazing products and experiences you will build with Meta Llama 3.

Our latest updates delivered to your inbox

Subscribe to our newsletter to keep up with Meta AI news, events, research breakthroughs, and more.

Join us in the pursuit of what’s possible with AI.

Product experiences

Foundational models

Latest news

Meta © 2024