Fastino Labs just released GLiGuard, an open-source safety moderation model that remembers encoders are king for these kinds of tasks.
One model, Apache 2.0:
gliguard-LLMGuardrails-300M: 300M params, evaluates multiple safety tasks at a time.
🧵
One model, Apache 2.0:
gliguard-LLMGuardrails-300M: 300M params, evaluates multiple safety tasks at a time.
🧵
May 13, 2026 at 2:41 PM
Fastino Labs just released GLiGuard, an open-source safety moderation model that remembers encoders are king for these kinds of tasks.
One model, Apache 2.0:
gliguard-LLMGuardrails-300M: 300M params, evaluates multiple safety tasks at a time.
🧵
One model, Apache 2.0:
gliguard-LLMGuardrails-300M: 300M params, evaluates multiple safety tasks at a time.
🧵
Across nine safety benchmarks, GLiGuard scores 87.7 F1 on prompt classification (within 1.7 of the best model) and 82.7 on response classification (second only to Qwen3Guard-8B).
It beats LlamaGuard4-12B, ShieldGemma-27B, and NemoGuard-8B at 23-90x smaller.
It beats LlamaGuard4-12B, ShieldGemma-27B, and NemoGuard-8B at 23-90x smaller.
May 13, 2026 at 2:41 PM
Across nine safety benchmarks, GLiGuard scores 87.7 F1 on prompt classification (within 1.7 of the best model) and 82.7 on response classification (second only to Qwen3Guard-8B).
It beats LlamaGuard4-12B, ShieldGemma-27B, and NemoGuard-8B at 23-90x smaller.
It beats LlamaGuard4-12B, ShieldGemma-27B, and NemoGuard-8B at 23-90x smaller.
Notion AI Agents, GLiGuard Open-Source Release, and OpenAI Supply Chain Response
#ai #artificialintelligence #future #machinelearning #ainews
#ai #artificialintelligence #future #machinelearning #ainews
May 14, 2026 at 3:40 PM
Notion AI Agents, GLiGuard Open-Source Release, and OpenAI Supply Chain Response
#ai #artificialintelligence #future #machinelearning #ainews
#ai #artificialintelligence #future #machinelearning #ainews
🤖 Company behind GLiNER model released open source model for running LLM guardrail
Fastino Labs has released GLiGuard, an open-source small language model designed for LLM safety moderati...
https://is.gd/gWUZkY #AINews #MachineLearning #CrustyTLDR
Fastino Labs has released GLiGuard, an open-source small language model designed for LLM safety moderati...
https://is.gd/gWUZkY #AINews #MachineLearning #CrustyTLDR
May 13, 2026 at 2:02 AM
🤖 Company behind GLiNER model released open source model for running LLM guardrail
Fastino Labs has released GLiGuard, an open-source small language model designed for LLM safety moderati...
https://is.gd/gWUZkY #AINews #MachineLearning #CrustyTLDR
Fastino Labs has released GLiGuard, an open-source small language model designed for LLM safety moderati...
https://is.gd/gWUZkY #AINews #MachineLearning #CrustyTLDR
Runs on a single GPU. Practical for teams that want guardrails in front of every input and every output without paying the latency tax of a 7B+ decoder on each check.
Model: huggingface.co/fastino/glig...
Model: huggingface.co/fastino/glig...
fastino/gliguard-LLMGuardrails-300M · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
May 13, 2026 at 2:41 PM
Runs on a single GPU. Practical for teams that want guardrails in front of every input and every output without paying the latency tax of a 7B+ decoder on each check.
Model: huggingface.co/fastino/glig...
Model: huggingface.co/fastino/glig...
And their blogpost: pioneer.ai/blog/gliguar...
Great work to Mary Newhauser, Urchade Zaratiana, and team! Big fan of the GLi-architectures ever since GLiNER back in the day.
Great work to Mary Newhauser, Urchade Zaratiana, and team! Big fan of the GLi-architectures ever since GLiNER back in the day.
GLiGuard: 16x Faster Safety Moderation with a Small Language Model - Pioneer AI by Fastino Labs
The agent building you a better model.
Adaptive Inference that continuously improves at runtime.
pioneer.ai
May 13, 2026 at 2:41 PM
And their blogpost: pioneer.ai/blog/gliguar...
Great work to Mary Newhauser, Urchade Zaratiana, and team! Big fan of the GLi-architectures ever since GLiNER back in the day.
Great work to Mary Newhauser, Urchade Zaratiana, and team! Big fan of the GLi-architectures ever since GLiNER back in the day.
GLiNER Guard (GLiGuard): один schema-driven энкодер вместо зоопарка LLM-гардрейлов Деплоите LLM? Значит, обвешиваете её гардами....
#GLiNER #Guard #GLiGuard #GLiNER #2 #guardrails #PII #zero-shot #безопасность #LLM #обработка
Origin | Interest | Match
#GLiNER #Guard #GLiGuard #GLiNER #2 #guardrails #PII #zero-shot #безопасность #LLM #обработка
Origin | Interest | Match
May 19, 2026 at 11:27 PM
GLiNER Guard (GLiGuard): один schema-driven энкодер вместо зоопарка LLM-гардрейлов Деплоите LLM? Значит, обвешиваете её гардами....
#GLiNER #Guard #GLiGuard #GLiNER #2 #guardrails #PII #zero-shot #безопасность #LLM #обработка
Origin | Interest | Match
#GLiNER #Guard #GLiGuard #GLiNER #2 #guardrails #PII #zero-shot #безопасность #LLM #обработка
Origin | Interest | Match
Most SOTA guardrails (LlamaGuard4, ShieldGemma, NemoGuard, WildGuard, Qwen3Guard, PolyGuard) are 7B-27B decoder LLMs that autoregressively generate verdicts token by token.
GLiGuard is a 300M encoder built on GLiNER2 that scores every label in one pass.
GLiGuard is a 300M encoder built on GLiNER2 that scores every label in one pass.
May 13, 2026 at 2:41 PM
Most SOTA guardrails (LlamaGuard4, ShieldGemma, NemoGuard, WildGuard, Qwen3Guard, PolyGuard) are 7B-27B decoder LLMs that autoregressively generate verdicts token by token.
GLiGuard is a 300M encoder built on GLiNER2 that scores every label in one pass.
GLiGuard is a 300M encoder built on GLiNER2 that scores every label in one pass.
Fastino Labs Unveils Innovative Language Models That Revolutionize AI Efficiency#United_States#Palo_Alto#Fastino_Labs#GLiNER2-PII#GLiGuard
Fastino Labs Unveils Innovative Language Models That Revolutionize AI Efficiency
Fastino Labs has launched GLiGuard and GLiNER2-PII, two groundbreaking language models that are significantly smaller and faster than existing alternatives while enhancing accuracy and efficiency.
third-news.com
May 14, 2026 at 1:30 PM
Fastino Labs Unveils Innovative Language Models That Revolutionize AI Efficiency#United_States#Palo_Alto#Fastino_Labs#GLiNER2-PII#GLiGuard
Model GLiGuard o rozmiarze 300 milionów parametrów działa do 16 razy szybciej niż rynkowi liderzy, oferując kompleksową ochronę w jednym cyklu obliczeniowym. To zmienia paradygmat bezpieczeństwa AI, czyniąc je błyskawiczną kategoryzacją danych, a nie generatywnym procesem.
Koniec absurdalnych opóźnień: GLiGuard udowadnia, że bezpieczeństwo AI nie wymaga miliardów parametrów
Model GLiGuard o rozmiarze 300 milionów parametrów działa do 16 razy szybciej niż rynkowi liderzy, oferując kompleksową ochronę w jednym cyklu obliczeniowym. To zmienia paradygmat bezpieczeństwa AI, czyniąc je błyskawiczną kategoryzacją danych, a nie generatywnym procesem.
aisight.pl
May 15, 2026 at 3:03 PM
Model GLiGuard o rozmiarze 300 milionów parametrów działa do 16 razy szybciej niż rynkowi liderzy, oferując kompleksową ochronę w jednym cyklu obliczeniowym. To zmienia paradygmat bezpieczeństwa AI, czyniąc je błyskawiczną kategoryzacją danych, a nie generatywnym procesem.
🧠 A company has released an open source model designed to run LLM guardrails. The model, called GLiNER, is now available for public use.
💬 Hacker News
🔗 https://pioneer.ai/blog/gliguard-16x-faster-safety-moderation-with-a-small-language-model
#AI #tech
💬 Hacker News
🔗 https://pioneer.ai/blog/gliguard-16x-faster-safety-moderation-with-a-small-language-model
#AI #tech
May 12, 2026 at 6:49 PM
🧠 A company has released an open source model designed to run LLM guardrails. The model, called GLiNER, is now available for public use.
💬 Hacker News
🔗 https://pioneer.ai/blog/gliguard-16x-faster-safety-moderation-with-a-small-language-model
#AI #tech
💬 Hacker News
🔗 https://pioneer.ai/blog/gliguard-16x-faster-safety-moderation-with-a-small-language-model
#AI #tech
Urchade Zaratiana, Mary Newhauser, George Hurn-Maloney, Ash Lewis: GLiGuard: Schema-Conditioned Classification for LLM Safeguard https://arxiv.org/abs/2605.07982 https://arxiv.org/pdf/2605.07982 https://arxiv.org/html/2605.07982
May 12, 2026 at 6:40 AM
Urchade Zaratiana, Mary Newhauser, George Hurn-Maloney, Ash Lewis: GLiGuard: Schema-Conditioned Classification for LLM Safeguard https://arxiv.org/abs/2605.07982 https://arxiv.org/pdf/2605.07982 https://arxiv.org/html/2605.07982
Urchade Zaratiana, Mary Newhauser, George Hurn-Maloney, Ash Lewis: GLiGuard: Schema-Conditioned Classification for LLM Safeguard https://arxiv.org/abs/2605.07982 https://arxiv.org/pdf/2605.07982 https://arxiv.org/html/2605.07982
May 11, 2026 at 6:41 AM
Urchade Zaratiana, Mary Newhauser, George Hurn-Maloney, Ash Lewis: GLiGuard: Schema-Conditioned Classification for LLM Safeguard https://arxiv.org/abs/2605.07982 https://arxiv.org/pdf/2605.07982 https://arxiv.org/html/2605.07982
GLiGuard: 16 Kat Daha Hızlı Güvenlik Moderasyonu İçin Yeni Model
Yapay zeka uygulamalarındaki güvenlik önlemleri, sistemlerin yaygınlaşmasıyla birlikte giderek daha kritik ve maliyetli hale geliyor. Özellikle büyük dil modelleri (LLM) üzerinden kullanıcıların gönderdiği mesajlar ile sistemin…
Yapay zeka uygulamalarındaki güvenlik önlemleri, sistemlerin yaygınlaşmasıyla birlikte giderek daha kritik ve maliyetli hale geliyor. Özellikle büyük dil modelleri (LLM) üzerinden kullanıcıların gönderdiği mesajlar ile sistemin…
GLiGuard: 16 Kat Daha Hızlı Güvenlik Moderasyonu İçin Yeni Model
Yapay zeka uygulamalarındaki güvenlik önlemleri, sistemlerin yaygınlaşmasıyla birlikte giderek daha kritik ve maliyetli hale geliyor. Özellikle büyük dil modelleri (LLM) üzerinden kullanıcıların gönderdiği mesajlar ile sistemin verdiği yanıtların değerlendirilmesi, gerçek zamanlı ve güvenilir bir moderasyon gerektiriyor. Ancak bugüne kadar kullanılan modeller ne yazık ki yüksek parametre sayılarına sahip ve yavaş çalıştığı için bu süreç ağır bir yük halini alıyordu. Bu soruna çözüm olarak Fastino Labs tarafından geliştirilen GLiGuard, dikkat çekici bir yenilik sunuyor.
incebilim.com
May 14, 2026 at 8:30 AM
GLiGuard: 16 Kat Daha Hızlı Güvenlik Moderasyonu İçin Yeni Model
Yapay zeka uygulamalarındaki güvenlik önlemleri, sistemlerin yaygınlaşmasıyla birlikte giderek daha kritik ve maliyetli hale geliyor. Özellikle büyük dil modelleri (LLM) üzerinden kullanıcıların gönderdiği mesajlar ile sistemin…
Yapay zeka uygulamalarındaki güvenlik önlemleri, sistemlerin yaygınlaşmasıyla birlikte giderek daha kritik ve maliyetli hale geliyor. Özellikle büyük dil modelleri (LLM) üzerinden kullanıcıların gönderdiği mesajlar ile sistemin…