{"article":{"annotation":{"article_id":"meta-opportunistic-inference-capacity-ads-recommendations","available_at":"2026-10-15T22:34:10.376345Z","geography_revision":"ef6ee0010dda5fe7b240fd514749b43e214c57da73222d2e40ec83b276dd0977","id":"04a464b2df942941ae0522ea77d315f1d3abadcc64deaa883807621d3d0a5522","input_sha256":"dfd0302d15c62df99645b6d4cbfb61bdb8d594fc5c374e44fa8bb2b4d9c5cb0f","limitations":["Rule-derived subjects are candidates, not verified editorial labels or calibrated model probabilities.","Coverage is distinct from event location. Unknown country/city/coordinates remain absent; no automatic geocoding.","Article geometries keep their stated precision and methodology; reproducing them is not independent coordinate verification."],"method":"controlled-rules-v1","places":[],"rules_sha256":"fb5ad3c27c91cb840f7a0a05f7f64a5cb26ad0374409ff9457f277c7092e22b4","schema_version":1,"subjects":[{"concept_id":"medtop:13000000","confidence":null,"confidence_method":null,"evidence":[{"field":"category+topics","language":"en","note":"Preserved editorial category/topic; broad candidate only","source_urls":["https://engineering.fb.com/2026/10/12/ml-applications/turning-opportunistic-ai-inference-capacity-into-better-ads-recommendations/"]}],"method":"controlled-rules-v1","rule_id":"technology-broad","status":"candidate"}],"taxonomy_revision":"1f2db31d811fca07e99b16d402c10e2cc2965c9df1c7b41d9be348b40a24e7fb","taxonomy_version":"IPTC Media Topics 2026-Q2"},"api_url":"https://newsverum.com/api/v1/articles/meta-opportunistic-inference-capacity-ads-recommendations?lang=en","category":"technology","content":{"key_points":[],"paragraphs":["Meta combines locally idle pockets across regions and accelerator types into a shared, opportunistic inference pool, even while the fleet is constrained overall. Its ads-ranking system uses two elasticity levers: Rank More Ads evaluates additional candidates, while Serve Better Models changes the share of requests sent to a more powerful model.","Rank More Ads needs surplus CPU and accelerator capacity for additional candidates. Serve Better Models keeps candidate volume fixed and routes selected requests to the stronger model, allowing it to use accelerator-only surplus. The two paths can consume different mixes of spare resources.","Meta isolates baseline serving from optional work physically or logically. Central control loops adjust over minutes, while server- and client-side load shedding can drop lower-value optional work within milliseconds if capacity is reclaimed.","The request router selects which requests receive the stronger model through an explore-and-exploit process that periodically measures incremental value; when capacity shrinks, those requests can fall back to the baseline model. Meta’s post details architecture and controls but gives no quantified uplift in ad performance."],"sections":[]},"corrections":[],"country":null,"data":{"blocks":[{"description":{"ar":"تجمع Meta الفجوات الخاملة محلياً عبر المناطق وأنواع المسرّعات في حوض استدلال مشترك وانتهازي، رغم أن الأسطول مقيد إجمالاً. ويستخدم نظام ترتيب الإعلانات أداتين للمرونة: Rank More Ads لتقييم مرشحين إضافيين، وServe Better Models لتغيير نسبة الطلبات الموجهة إلى نموذج أقوى.","de":"Meta bündelt lokal freie Kapazität über Regionen und Beschleunigertypen hinweg zu einem gemeinsamen, opportunistischen Inferenzpool, obwohl die Flotte insgesamt ausgelastet sein kann. Das Anzeigensystem nutzt zwei Stellhebel: Rank More Ads bewertet zusätzliche Kandidaten; Serve Better Models erhöht oder senkt den Anteil der Anfragen an ein leistungsfähigeres Modell.","en":"Meta combines locally idle pockets across regions and accelerator types into a shared, opportunistic inference pool, even while the fleet is constrained overall. Its ads-ranking system uses two elasticity levers: Rank More Ads evaluates additional candidates, while Serve Better Models changes the share of requests sent to a more powerful model.","es":"Meta agrupa los espacios de capacidad ociosa por región y tipo de acelerador en un fondo compartido de inferencia oportunista, aunque la flota esté limitada en conjunto. Su sistema de clasificación de anuncios usa dos palancas: Rank More Ads evalúa candidatos adicionales y Serve Better Models cambia la proporción de solicitudes que se envía a un modelo más potente.","fr":"Meta regroupe les poches de capacité inutilisée selon les régions et les accélérateurs dans un pool d’inférence opportuniste partagé, même lorsque la flotte est globalement contrainte. Son système de classement publicitaire utilise deux leviers: Rank More Ads évalue davantage de candidats, tandis que Serve Better Models modifie la part des requêtes dirigées vers un modèle plus puissant.","pt":"A Meta reúne bolsas de capacidade ociosa por região e tipo de acelerador num conjunto partilhado de inferência oportunista, mesmo quando a frota está globalmente limitada. O sistema de classificação de anúncios usa duas alavancas: Rank More Ads avalia candidatos adicionais e Serve Better Models altera a proporção de pedidos enviada para um modelo mais potente.","tr":"Meta, küresel ölçekte kısıtlı olsa da bölge ve hızlandırıcı türlerine göre yerel boşluklar içeren çıkarım filosunu ortak bir isteğe bağlı kapasite havuzuna bağlıyor. Sistem, reklam sıralamasında iki esneklik kullanıyor: Rank More Ads daha fazla adayı değerlendiriyor; Serve Better Models ise isteklerin daha güçlü modele giden payını değiştiriyor."},"id":"source-event","items":[{"date":"2026-10-12","description":{"ar":"تجمع Meta الفجوات الخاملة محلياً عبر المناطق وأنواع المسرّعات في حوض استدلال مشترك وانتهازي، رغم أن الأسطول مقيد إجمالاً. ويستخدم نظام ترتيب الإعلانات أداتين للمرونة: Rank More Ads لتقييم مرشحين إضافيين، وServe Better Models لتغيير نسبة الطلبات الموجهة إلى نموذج أقوى.","de":"Meta bündelt lokal freie Kapazität über Regionen und Beschleunigertypen hinweg zu einem gemeinsamen, opportunistischen Inferenzpool, obwohl die Flotte insgesamt ausgelastet sein kann. Das Anzeigensystem nutzt zwei Stellhebel: Rank More Ads bewertet zusätzliche Kandidaten; Serve Better Models erhöht oder senkt den Anteil der Anfragen an ein leistungsfähigeres Modell.","en":"Meta combines locally idle pockets across regions and accelerator types into a shared, opportunistic inference pool, even while the fleet is constrained overall. Its ads-ranking system uses two elasticity levers: Rank More Ads evaluates additional candidates, while Serve Better Models changes the share of requests sent to a more powerful model.","es":"Meta agrupa los espacios de capacidad ociosa por región y tipo de acelerador en un fondo compartido de inferencia oportunista, aunque la flota esté limitada en conjunto. Su sistema de clasificación de anuncios usa dos palancas: Rank More Ads evalúa candidatos adicionales y Serve Better Models cambia la proporción de solicitudes que se envía a un modelo más potente.","fr":"Meta regroupe les poches de capacité inutilisée selon les régions et les accélérateurs dans un pool d’inférence opportuniste partagé, même lorsque la flotte est globalement contrainte. Son système de classement publicitaire utilise deux leviers: Rank More Ads évalue davantage de candidats, tandis que Serve Better Models modifie la part des requêtes dirigées vers un modèle plus puissant.","pt":"A Meta reúne bolsas de capacidade ociosa por região e tipo de acelerador num conjunto partilhado de inferência oportunista, mesmo quando a frota está globalmente limitada. O sistema de classificação de anúncios usa duas alavancas: Rank More Ads avalia candidatos adicionais e Serve Better Models altera a proporção de pedidos enviada para um modelo mais potente.","tr":"Meta, küresel ölçekte kısıtlı olsa da bölge ve hızlandırıcı türlerine göre yerel boşluklar içeren çıkarım filosunu ortak bir isteğe bağlı kapasite havuzuna bağlıyor. Sistem, reklam sıralamasında iki esneklik kullanıyor: Rank More Ads daha fazla adayı değerlendiriyor; Serve Better Models ise isteklerin daha güçlü modele giden payını değiştiriyor."},"label":{"ar":"Meta تستخدم سعة الاستدلال الخاملة لتحسين توصيات الإعلانات","de":"Meta nutzt freie Inferenzkapazität für Anzeigenempfehlungen","en":"Meta uses idle inference capacity for ads recommendations","es":"Meta aprovecha capacidad de inferencia ociosa para recomendar anuncios","fr":"Meta exploite la capacité d’inférence inutilisée pour recommander des publicités","pt":"A Meta usa capacidade de inferência ociosa para recomendações de anúncios","tr":"Meta, reklam önerileri için boşta kalan çıkarım kapasitesini kullanıyor"},"status":"occurred"}],"methodology":{"ar":"يلخص سجل الحدث هذا البيان المؤرخ في المصدر الأولي المحدد. ولا تُعرض ادعاءات المنتج أو الاختبار الواردة فيه على أنها نتائج تم التحقق منها بصورة مستقلة.","de":"Dieser Ereigniseintrag fasst die datierte Mitteilung der genannten Primärquelle zusammen. Produkt- oder Testangaben der Quelle werden nicht als unabhängig verifizierte Ergebnisse dargestellt.","en":"This event record summarizes the dated statement in the cited primary source. Product or test claims from the source are not presented as independently verified results.","es":"Este registro de evento resume la publicación fechada de la fuente primaria citada. Las afirmaciones de producto o prueba de la fuente no se presentan como resultados verificados de forma independiente.","fr":"Cette fiche d’événement résume la publication datée de la source primaire citée. Les affirmations de produit ou de test de la source ne sont pas présentées comme des résultats vérifiés indépendamment.","pt":"Este registo do evento resume a publicação datada da fonte primária citada. As alegações de produto ou teste da fonte não são apresentadas como resultados verificados de forma independente.","tr":"Bu olay kaydı, belirtilen birincil kaynağın tarihli açıklamasını özetler. Kaynakta verilen ürün veya test iddiaları bağımsız olarak doğrulanmış sonuçlar olarak sunulmaz."},"sourceUrls":["https://engineering.fb.com/2026/10/12/ml-applications/turning-opportunistic-ai-inference-capacity-into-better-ads-recommendations/"],"title":{"ar":"Meta تستخدم سعة الاستدلال الخاملة لتحسين توصيات الإعلانات","de":"Meta nutzt freie Inferenzkapazität für Anzeigenempfehlungen","en":"Meta uses idle inference capacity for ads recommendations","es":"Meta aprovecha capacidad de inferencia ociosa para recomendar anuncios","fr":"Meta exploite la capacité d’inférence inutilisée pour recommander des publicités","pt":"A Meta usa capacidade de inferência ociosa para recomendações de anúncios","tr":"Meta, reklam önerileri için boşta kalan çıkarım kapasitesini kullanıyor"},"type":"events","updatedAt":"2026-10-15T22:22:55+00:00"}],"version":1},"data_downloads":[{"formats":{"csv":"https://newsverum.com/api/v1/articles/meta-opportunistic-inference-capacity-ads-recommendations/data/source-event?lang=en&format=csv","json":"https://newsverum.com/api/v1/articles/meta-opportunistic-inference-capacity-ads-recommendations/data/source-event?lang=en&format=json"},"id":"source-event"}],"data_types":["events"],"editorial_illustration":{"alt":"Conceptual illustration of technology, processors and connected information.","asset_id":"newsverum-v1-technology","avif_variants":[{"height":320,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-480.avif","width":480},{"height":480,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-720.avif","width":720},{"height":640,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-960.avif","width":960},{"height":800,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1200.avif","width":1200},{"height":1024,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.avif","width":1536}],"credit":"Newsverum","documentary":false,"height":1024,"kind":"symbolic-editorial-illustration","language":"en","scope":"topic","url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.webp","variants":[{"height":320,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-480.webp","width":480},{"height":640,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-960.webp","width":960},{"height":1024,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.webp","width":1536}],"width":1536},"event_id":"event-1daeff21-a41f-5ba3-a387-b61d7ea148dc","event_period":"2026-10-12","event_period_basis":"event-or-reference-period","id":"meta-opportunistic-inference-capacity-ads-recommendations","image":null,"importance":0.6,"language":"en","modified_at":"2026-10-15T22:33:53+00:00","publication_basis":"publication-record","published_at":"2026-10-15T22:33:53+00:00","reading_minutes":2,"region":"world","relations":[],"section_illustration":{"alt":"Conceptual illustration of technology, processors and connected information.","asset_id":"newsverum-v1-technology","avif_variants":[{"height":320,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-480.avif","width":480},{"height":480,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-720.avif","width":720},{"height":640,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-960.avif","width":960},{"height":800,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1200.avif","width":1200},{"height":1024,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.avif","width":1536}],"credit":"Newsverum","documentary":false,"height":1024,"kind":"symbolic-editorial-illustration","language":"en","scope":"section","url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.webp","variants":[{"height":320,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-480.webp","width":480},{"height":640,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-960.webp","width":960},{"height":1024,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.webp","width":1536}],"width":1536},"slug":"meta-opportunistic-inference-capacity-ads-recommendations","sources":[{"checkedAt":"2026-10-15","name":"Meta Engineering","publishedAt":"2026-10-12","url":"https://engineering.fb.com/2026/10/12/ml-applications/turning-opportunistic-ai-inference-capacity-into-better-ads-recommendations/"}],"summary":"Meta combines locally idle pockets across regions and accelerator types into a shared, opportunistic inference pool, even while the fleet is constrained overall. Its ads-ranking system uses two elasticity levers: Rank More Ads evaluates additional candidates, while Serve Better Models changes the share of requests sent to a more powerful model.","text":"Meta combines locally idle pockets across regions and accelerator types into a shared, opportunistic inference pool, even while the fleet is constrained overall. Its ads-ranking system uses two elasticity levers: Rank More Ads evaluates additional candidates, while Serve Better Models changes the share of requests sent to a more powerful model.\n\nMeta combines locally idle pockets across regions and accelerator types into a shared, opportunistic inference pool, even while the fleet is constrained overall. Its ads-ranking system uses two elasticity levers: Rank More Ads evaluates additional candidates, while Serve Better Models changes the share of requests sent to a more powerful model.\n\nRank More Ads needs surplus CPU and accelerator capacity for additional candidates. Serve Better Models keeps candidate volume fixed and routes selected requests to the stronger model, allowing it to use accelerator-only surplus. The two paths can consume different mixes of spare resources.\n\nMeta isolates baseline serving from optional work physically or logically. Central control loops adjust over minutes, while server- and client-side load shedding can drop lower-value optional work within milliseconds if capacity is reclaimed.\n\nThe request router selects which requests receive the stronger model through an explore-and-exploit process that periodically measures incremental value; when capacity shrinks, those requests can fall back to the baseline model. Meta’s post details architecture and controls but gives no quantified uplift in ad performance.","title":"Meta uses idle inference capacity for ads recommendations","topics":[],"translations":[{"api_url":"https://newsverum.com/api/v1/articles/meta-opportunistic-inference-capacity-ads-recommendations?lang=tr","language":"tr","url":"https://newsverum.com/tr/news/meta-opportunistic-inference-capacity-ads-recommendations"},{"api_url":"https://newsverum.com/api/v1/articles/meta-opportunistic-inference-capacity-ads-recommendations?lang=en","language":"en","url":"https://newsverum.com/en/news/meta-opportunistic-inference-capacity-ads-recommendations"},{"api_url":"https://newsverum.com/api/v1/articles/meta-opportunistic-inference-capacity-ads-recommendations?lang=ar","language":"ar","url":"https://newsverum.com/ar/news/meta-opportunistic-inference-capacity-ads-recommendations"},{"api_url":"https://newsverum.com/api/v1/articles/meta-opportunistic-inference-capacity-ads-recommendations?lang=fr","language":"fr","url":"https://newsverum.com/fr/news/meta-opportunistic-inference-capacity-ads-recommendations"},{"api_url":"https://newsverum.com/api/v1/articles/meta-opportunistic-inference-capacity-ads-recommendations?lang=es","language":"es","url":"https://newsverum.com/es/news/meta-opportunistic-inference-capacity-ads-recommendations"},{"api_url":"https://newsverum.com/api/v1/articles/meta-opportunistic-inference-capacity-ads-recommendations?lang=de","language":"de","url":"https://newsverum.com/de/news/meta-opportunistic-inference-capacity-ads-recommendations"},{"api_url":"https://newsverum.com/api/v1/articles/meta-opportunistic-inference-capacity-ads-recommendations?lang=pt","language":"pt","url":"https://newsverum.com/pt/news/meta-opportunistic-inference-capacity-ads-recommendations"}],"url":"https://newsverum.com/en/news/meta-opportunistic-inference-capacity-ads-recommendations"},"version":"1"}