As of 2026-07-04, Google Gemma 3 1B has an attack success rate (ASR) of 60.0% on Guardion's LLM Vulnerability benchmark — ranking #92 of 93 models. It is highly vulnerable — most adversarial prompts succeed without an external guardrail.
Attack Success Rate (ASR) is the share of adversarial prompts that elicit harmful output across zero-shot, TAP, and Crescendo attacks (HarmBench framework). Lower is safer. "Est." models are calibrated from public safety evaluations pending a Guardion benchmark run.
Google Gemma 3 1B is highly vulnerable — most adversarial prompts succeed without an external guardrail. On Guardion's LLM Vulnerability benchmark it records a 60.0% attack success rate (40% robustness), ranking #92 of 93 models (estimated from public safety evaluations).
Gemma 3 1B's overall ASR is 60.0% — the share of adversarial prompts that elicit harmful output across zero-shot, TAP, and Crescendo attacks. Lower is safer.
Gemini 3 Pro is more robust: 6.0% ASR vs Gemma 3 1B's 60.0%. See the full head-to-head comparison for the per-attack breakdown.
Even robust models are bypassed under sustained attack. An inline runtime guardrail that classifies prompts and responses — like Guardion's prompt-defense models — blocks jailbreaks and prompt injection before they reach Gemma 3 1B.
Guardion's runtime guardrails block jailbreaks and prompt injection before they reach any model — with sub-130ms policy decisions.
Request a demo