AI Briefing
KOSign in

Axiomic Labs releases X3Guard-150M, a 150M parameter prompt guard model

·2026.11.16 00:46

Key point

The model runs in-browser with response times in the tens of milliseconds on iPhone and outperforms larger competitors.

Details

  • Axiomic Labs released X3Guard-150M, a 150M parameter prompt guard model trained on the GPT-X3 base model.
  • The model simultaneously flags whether a prompt is Harmful and whether it is Adversarial.
  • X3Guard-150M outperforms models many times its size while running entirely on edge devices.
  • In-browser response times are in the tens of milliseconds on iPhone.
llmsafetyedge-aiprompt-guardaxiomic-labs
Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.

Related

RedditNVIDIA releases Nemotron 3.5 Bolt E4B, a 4GB post-trained Gemma 4 model for edge devicesPyTorchKRLiquid AI Unveils Six 2B-Scale LFM3 ModelsRedditSalesforce releases SOV-Transcribe-1, an open-weight speech recognition modelPyTorchKRBotPsy-0.8B Achieves Higher Tool-Use Performance Than 2B ModelsTLDR AIStudy demonstrates LLMs can learn multi-step reasoning on filler tokensTLDR AIPyTorch ExecuTorch Adds WebGPU Backend for In-Browser GPU Inference