Meta Release Notes

We're excited to release Muse Code (beta), a terminal coding agent powered by Muse Spark 1.2, our newest model. This marks our next step toward the frontier, with larger and much more capable models on the way.

Install Muse Code on macOS or Linux:

curl -fsSL https://dev.meta.ai/install.sh | bash

Muse Code takes on complex software engineering tasks across large repositories: planning changes, writing code, and validating the results. It can coordinate multiple persistent subagents for each task, solving difficult problems faster, more accurately, and with less intervention.

Muse Code

Async Background Agents

Muse Code operates with a simple agent loop plus a set of async background agents to enhance the main agent's capability. These specialized background agents remain active throughout each session, rather than being spawned for individual tasks, helping avoid redundant information gathering. They carry out next steps and choose when to communicate back to the main agent. Their persistence reduces latency and the need for steering on difficult, multi-step tasks.

Runtime Design

Muse Code uses a local event log in which every model call, tool run, approval, and edit is appended. This single source of truth makes the runtime replay-exact and restart-safe: after a crash, the agent can resume precisely where it stopped. That ability lets Muse Code take on long-running tasks without being derailed by failures.

Bundled Skills

Muse Code ships with several default skills.

/plan turns a task into an approval-gated plan,

/grill stress-tests that plan until it holds up, and

/goal works toward successful completion of the specified objective.

Muse Code rebuilds a luxury vacation rental booking page from a reference video and validates the finished site in a browser. The user inputs a fly-through video of a home into the terminal as an mp4 file. Muse Code interprets the video and produces a visually rich vacation home marketing and booking page.

Muse Spark 1.2

Muse Spark 1.2 is a coding-focused update to Muse Spark 1.1, with improvements in code generation, complex debugging, codebase understanding, and end-to-end developer workflows. In Muse Spark 1.2, we significantly scaled up training compute on coding tasks while expanding training environment diversity. The model also maintains its strength in other key areas like general agents.

For more details about our evaluations, see our report.

Co-Training With Muse Code

We co-trained Muse Spark 1.2 with Muse Code to ensure the model exhibits its best performance and coding usability when paired together. The training included rejection sampled harness trajectories and recipe optimizations for goals, compaction, and subagents, alongside the integration of the Muse Code toolset to maximize harness compatibility.

Long-Horizon

Muse Spark 1.2 was extensively trained on long-horizon coding tasks, including whole-repository generation, large end-to-end projects, and auto-research. It leverages planning to sequence work, goal conditioning to maintain direction, and context compaction to retain the knowledge needed to sustain progress.

Self-Improvement

We also used Muse Spark 1.1 to generate challenging coding environments and instruction-following templates. The model then graded candidate solutions on how well they satisfied those requirements, producing a scalable training dataset for Muse Spark 1.2. This self-improvement loop helped Muse Spark 1.2 follow complex instructions more precisely than its predecessor.

Case Study: Kernel Optimization

We tested the model's ability to iteratively optimize GPU kernels over 1,000+ tool calls (up to 24 hours). Leveraging Muse Code's agentic coding environment, the model writes, compiles, profiles, and progressively improves kernel performance relative to a provided baseline implementation. We benchmarked on KDA and MLA kernels for NVIDIA Hopper GPUs. The agent continues to achieve substantial improvements over the provided baseline implementation.

The baseline is the FLA Triton implementation of KDA. Models were prohibited from importing third-party kernel libraries such as FLA directly; instead, they had to apply specialized kernel-optimization knowledge to implement the algorithm in Triton, rather than wrap existing implementations. Muse Spark 1.2 paired a chunk-parallel preparation kernel with a sequential inter-chunk scan, combining standard fusion and tiling with KDA-specific optimizations such as re-centering the gated cumulative decay at the chunk midpoint.

Availability

Muse Spark 1.2 is available today in Muse Code and in Meta Model API with expanded global access. We have a lot on the horizon, including new harness features and more powerful models. We can’t wait to see what you build!

Get started with Muse Code

Original source

Group chats are best on WhatsApp – from staying close to your inner circle to coordinating with your fantasy league. Today we’re making them even better with new polls updates for faster decision-making, @all mentions to get everyone’s attention, and a simpler way to create new group chats from existing groups.

Better polls:

We're making polls even more useful for group decisions with three new improvements. You can now set an end time to lock voting when time is up, hide voter names so the whole group is more comfortable sharing their preferences, and edit your poll questions within 15 minutes if you spot a typo or need to clarify. Whether you're finalizing when or where to meet, it's now easier to quickly go from question to decision.

@all:

Alert everyone in your group chat about important, time-sensitive messages. No more tagging people one by one or hoping your message gets noticed in a busy group chat. Use @all to notify everyone about urgent messages, like school or office closures, event deadlines, or last-minute changes to your plans. For larger groups with more than 32 people, @all is available for admins only.

While the @all feature will alert you to important messages even in muted group chats, you remain in control of how you receive notifications and can mute @all at any time in notification settings.

Create group chats from existing groups:

Instantly start a new group chat with people from an existing one in just a tap – no need to add people one by one. It's perfect for splitting off into a side conversation to plan a surprise party, coordinate event logistics, or dive deeper into a topic without cluttering the main chat.

Earlier this year, we upgraded group chats with the launch of group message history, member tags and event reminders. We're continuing that momentum with smarter polls, @all mentions, and faster group chat creation. With more than 100 million people in the U.S. on WhatsApp, and over 3 billion people globally, there's no better place for your group chats.

Original source

Takeaways

• We’re introducing new updates to WhatsApp group chats, including improvements to polls, @all mentions to get the whole chat’s attention, and the ability to create chats from existing groups.

Group chats are best on WhatsApp, helping you connect with groups you care about, from your inner circle to your fantasy football league. Today we’re making group chats even better with updates to polls for faster decision-making, @all mentions to get everyone’s attention, and the ability to create new group chats from existing ones.

Coordinate With Your Group

We’re making polls even more useful for group decisions with three new improvements. You can now set an end time to lock voting when time is up, hide voter names so the whole group is more comfortable sharing their preferences, and edit your poll questions within 15 minutes if you spot a typo or need to make a clarification. Whether you’re finalizing when to meet, where to meet, or what to do when you get there, it’s now easier to quickly go from question to decision.

Alert the Whole Chat

No more tagging people one by one or hoping your message gets noticed in a busy group chat. Use @all to notify everyone about urgent messages, like school or office closures, event deadlines, or last-minute changes to your plans. For larger groups with more than 32 people, @all is available for admins only.

While the @all feature will alert you to important messages even in muted group chats, you remain in control of how you receive notifications and can mute @all at any time in notification settings.

Create Chats From Existing Groups

You can now start a new group chat with people from an existing one in just a tap — no need to add people one by one. It’s perfect for splitting off into a side conversation to plan a surprise party, coordinate event logistics, or dive deeper into a topic without cluttering the main chat.

Earlier this year, we upgraded group chats with the launch of group message history, member tags and event reminders. We’re continuing that momentum with today’s updates. WhatsApp already offers the best group chat experience, and we’re committed to making it even better.

Original source

Download and copyright protections on media_url

Download and copyright protections on media_url

Applies to all versions.

The media_url field on the IG Media endpoint is now omitted from responses in either of these cases:

• The video contains copyrighted or licensed audio, including audio added from the Instagram audio library, or it has been flagged for a copyright violation. This applies to all video media, including reels, video posts, and video stories, and applies even when the app user owns the media.
• The reel owner has turned off downloads for the reel, through either the account-level or per-reel Allow people to download your reels setting. This applies to reels only, and only on the endpoints listed below that read other users’ media. The setting is evaluated against the reel’s owner, so it can apply to a reel you co-authored. Reading your own media through GET //media or GET / is not affected by this condition.

This gives media returned by the Instagram APIs the same download and copyright protections that already apply in the Instagram app. All other fields on the media object, such as id, permalink, caption, and timestamp, are returned as normal. Image media and carousel containers are not affected, though a carousel’s video children can be. Requests still return a 200 response. Treat media_url as an optional field in your app and fall back to permalink or thumbnail_url when it is absent.

This change applies to both graph.facebook.com and graph.instagram.com.

Endpoints affected by both the download setting and copyright checks

• GET /?fields=business_discovery{media}
• GET //tags
• GET //mentioned_media
• GET //mentioned_comment
• GET //top_media
• GET //recent_media
• GET //collaborative_media

Endpoints affected by the copyright check only

• GET /
• GET //children
• GET //media
• GET //stories
• GET //live_media
• GET //collaboration_invites

Nested media reached through field expansion, such as carousel children or the media on a comment, inherit the conditions that apply to the originating request. For example, GET //children is subject to the copyright check alone, but children returned by business_discovery{media{children}} are also subject to the download setting.

For more information, see Allow people to download your reels⁠ This link opens in a new tab or window, Copyright⁠ This link opens in a new tab or window, and Using audio in reels⁠ This link opens in a new tab or window.

Original source

Ads in WhatsApp Status are available via the Marketing API.

Learn more about ads in WhatsApp Status.

Marketing API

July 29, 2026

Available until TBD

Blog post

Ad Creative

WhatsApp Status ads: wamo_whatsapp_identity_spec

Applies to v26.0+.

The wamo_whatsapp_identity_spec field is now available on ad creative creation and readable on ad creative GET endpoints for GK-enabled ad accounts. This field allows advertisers to specify a WhatsApp identity when creating creatives for WhatsApp Status ad placements.

Additionally, if the WhatsApp Status placement is included and user_age_unknown is not set, it will default to true.

The following endpoints are affected:

• POST /{ad-account-id}/adcreatives
• GET /{ad-creative-id}
• GET /{ad-account-id}/adcreatives
• GET /{campaign-id}/adcreatives
• GET /{ad-id}/adcreatives

Shops Ads default to WEBSITE_AND_SHOP

Applies to v26.0+.

Eligible Shops Ads creatives will now automatically default to destination_spec.destination_type = WEBSITE_AND_SHOP when the advertiser has an IAB offsite checkout-enabled shop. Advertisers can opt out by explicitly setting destination_type to WEBSITE_AND_SHOP_OPT_OUT.

The following endpoint is affected:

• POST /{ad-account-id}/adcreatives

Poll ads deprecation

Applies to v26.0+. Will apply to all versions October 27, 2026.

Poll is no longer available in ad creative. The poll_spec field and the "poll" enum value for the type field under interactive_components_spec are no longer available when creating or updating ads. Poll-related fields should be removed from the creative spec.

The following endpoints are affected:

• POST /{ad-account-id}/adcreatives
• POST /{ad-account-id}/ads
• POST /{ad-id}
• GET /{ad-account-id}/generatepreviews

Ads

applink_treatment=web_only deprecation for Web and App campaigns

Applies to v26.0+. Will apply to all versions October 27, 2026.

The applink_treatment field set to web_only is deprecated for Web and App campaigns. Use automatic or deeplink_with_web_fallback instead.

The following endpoints are affected:

• POST /act_{ad-account-id}/ads
• POST /{ad-id}

Advantage+ Campaigns

Explicit advantage_audience flag required for HEF campaigns

Applies to v26.0+. Will apply to all versions October 27, 2026.

Campaigns with Special Ad Categories for Housing, Employment, or Financial products and services (HEF) that use targeting automation via the Marketing API must now explicitly set the advantage_audience flag. Previously, HEF campaigns could create ad sets with constrained targeting setups without specifying this flag.

With this change, if a HEF campaign has constrained targeting (for example, custom audience inclusions or detailed targeting), the API will return an error when the advantage_audience flag is not explicitly provided. This aligns HEF campaign behavior with the existing requirement for non-HEF campaigns and ensures advertisers are making an intentional choice about Advantage+ Audience for regulated ad categories.

The following endpoints are affected:

• POST /{ad-account-id}/adsets
• POST /{adset-id}/copies

Audiences

Lookalike country and location_spec deprecation

Applies to v26.0+. Will apply to all versions September 1, 2026.

The country and location_spec fields in lookalike_spec are deprecated when creating Lookalike audiences. These fields are now silently ignored and the API creates a parent (countryless) Lookalike audience. Location targeting is determined by the ad set, matching existing Audience Manager behavior. Callers passing deprecated fields will receive a deprecation warning but the request will succeed.

When country is specified on new Lookalikes, advertisers will see the following message:

The country and location_spec fields are no longer available in lookalike_spec. Your lookalike audiences will deliver to the locations selected in your ad sets’ audiences.

The following endpoint is affected:

• POST /{ad-account-id}/customaudiences

Insights

Delivery Estimate: daily_outcomes_curve, budget_guardrail, estimate_dau deprecation

Applies to v26.0+. Will apply to all versions October 27, 2026.

The daily_outcomes_curve, budget_guardrail, and estimate_dau fields on the Delivery Estimate endpoint were originally powered by the Bison / Outcome Prediction backend, which has been fully deprecated. These fields currently return hardcoded 0 or empty data and provide no useful information to developers, so they are being removed.

The following endpoints are affected:

• GET /{ad-account-id}/delivery_estimate
• GET /{adset-id}/delivery_estimate

Placements

Instagram Explore Feed placement removed

Applies to v26.0+.

The Instagram Explore Feed placement is no longer available.

The following endpoints are affected:

• POST /act_{ad-account-id}/adsets
• POST /{ad_set-id}

Messenger Stories placement removed

Applies to v26.0+. Will apply to all versions late August 2026.

The Messenger Stories placement (story in messenger_positions) is no longer available. Campaigns using this placement will have it automatically removed from their effective placements.

Migration guidance: Remove "story" from the messenger_positions array in your targeting spec. If messenger_positions only contained "story", remove the messenger_positions field entirely. If you are using Advantage+ Placements (automatic placements), no action is needed — the system will automatically exclude this placement.

The following endpoints are affected:

• POST /act_{ad-account-id}/adsets
• POST /{ad_set-id}

Did you find this page helpful?

Original source

Whether you're on your phone, at your desk, or switching between the two, calling on WhatsApp just works, privately and simply. Today, we are bringing new features to make that easier no matter where you are or what device you're on.

Web Calling

You can now make and receive audio and video calls, both one-on-one and group, directly from WhatsApp Web. No app download required.

Whether you're a college student on a shared computer, using a work laptop that doesn't allow app downloads, or simply prefer your browser, you can jump on a call without leaving the browser you're already on. You'll have access to features available on your other devices including screen sharing, reactions, and a dedicated Calls tab with your full call history and favorites, all from your browser.

Your calls on the web are end-to-end encrypted — with no time limits and at no cost. It’s the same private experience you expect from WhatsApp.

Call Transfer

Move an active group call from one device to another without hanging up. For those moments when you’re on the go and start a call on your mobile device or tablet, you can seamlessly transfer it to WhatsApp Web or Desktop when you arrive home to collaborate on a larger screen – or vice versa.

Waiting Room

For when you want more control of who joins your group calls and when, we've introduced Waiting Room. When you create a WhatsApp call link with "Require approval to join" enabled, participants will enter a waiting room until you're ready to let them in.

QuickHD

We’ve been improving the video experience at the beginning of a call. With QuickHD, you can now enjoy high-definition video immediately in the very first few seconds of the call.

Noise Suppression

We've launched noise suppression to remove the background noise around you so your voice comes through clearly to the person you’re calling – even in loud or busy environments. It can be managed at any time in your in-call settings.

These features are rolling out gradually and will be available to everyone soon. We hope they make your WhatsApp calling even simpler to use.

Original source

Takeaways

• We’re introducing new features to make WhatsApp calls even easier, like calling from WhatsApp Web, call transfer, background noise suppression, and more.

Whether you’re on your phone, at your desk, or switching between the two, WhatsApp gives you a private and simple way to make calls. Today, we are bringing new features to make calling even easier, no matter where you are or what device you’re on.

Call Directly From WhatsApp Web

You can now make and receive audio and video calls, both one-on-one and group, directly from WhatsApp Web. No app download required.

Whether you’re a college student on a shared computer, using a work laptop that doesn’t allow app downloads, or simply prefer your browser, you can jump on a call without leaving the browser you’re already on. You’ll have access to features available on your other devices including screen sharing, reactions, and a dedicated Calls tab with your full call history and favorites, all from your browser.

Your calls on the web are end-to-end encrypted — with no time limits and at no cost. It’s the same private experience you expect from WhatsApp.

Transfer Calls Between Devices

Move an active group call from one device to another without hanging up. For those moments when you’re on the go and start a call on your mobile device or tablet, you can seamlessly transfer it to WhatsApp Web or Desktop when you arrive home to collaborate on a larger screen – or vice versa.

A Waiting Room for Group Calls

For when you want more control of who joins your group calls and when, we’ve introduced waiting room. When you create a WhatsApp call link with “Require approval to join” enabled, participants will enter a waiting room until you’re ready to let them in.

Instant HD Video

We’ve been improving the video experience at the beginning of a call. With QuickHD, you can now enjoy high-definition video immediately in the very first few seconds of the call.

Drown Out the Noise

We’ve launched noise suppression to remove the background noise around you so your voice comes through clearly to the person you’re calling — even in loud or busy environments. It can be managed at any time in your in-call settings.

These features are rolling out gradually and will be available to everyone soon. We hope they make your WhatsApp calling even simpler to use.

Original source