SAI LabsBook a demo

Safety IndexGoogleGemini 3 Pro

Gemini 3 Pro

GoogleProprietaryReleased 2025-11Reasoning model

SAI Safety Index54provisionalHas 2 of the 4 pillars needed
Strongest pillar66Refusal calibration
Weakest pillar43Interaction and honesty
Coverage2of 5 pillars3 counted measurements, 0 by SAI

Summary

Gemini 3 Pro has data in 2 of the five pillars, short of the 4 needed for a headline number, so it is listed as partly measured; the mean of its measured pillars is 54. Its strongest area is refusal calibration at 66, and its weakest is interaction and honesty at 43.

3 measurements count toward the score, 0 of them from SAI's own runs. Not yet measured: adversarial robustness, agentic safety, people and society.

Pillar scores

How Gemini 3 Pro compares with every other model scored in the pillar, and which sub-categories the score rests on.

Refusal calibration

Does the model refuse what it should and help with what it may?

  • Harmful requests66 from 2 evaluations
  • Over-refusalnot measured

Rank 32 of 39 scored models.

Adversarial robustness

Do its refusals and instructions hold up when someone tries to get around them?

Not measured for Gemini 3 Pro yet.
  • Jailbreaksnot measured
  • Prompt injectionnot measured
  • Multilingual safetynot measured

9 models scored.

Agentic safety

Does it behave safely when it has tools and autonomy?

Not measured for Gemini 3 Pro yet.
  • Agent misusenot measured
  • Reward hacking and task integritynot measured
  • Overreach and destructive actionsnot measured

3 models scored.

Interaction and honesty

Is it honest with the person it is talking to?

  • Factuality and hallucinationnot measured
  • Honesty under pressure and sycophancy43 from 1 evaluation
  • Deception and schemingnot measured
  • Alignment auditnot measured
  • Manipulation of the usernot measured

Rank 26 of 29 scored models.

People and society

Does it protect vulnerable people and the public interest?

Not measured for Gemini 3 Pro yet.
  • Self-harm and wellbeingnot measured
  • Child safetynot measured
  • Bias, fairness and neutralitynot measured
  • Privacy: memorisation and leakagenot measured
  • Privacy: surveillance assistancenot measured
  • Elections and influence operationsnot measured

3 models scored.

Every measurement

These are the raw numbers behind the scores, with who measured them, when, and how. Rows marked counted feed the index. The others are shown for context, such as benign twins of harm rows, retired evaluations that every frontier model passes, tests of dangerous knowledge, other evaluators' overall scores, and the developer's own figures.

EvaluationValueScoreSourceSetting
Refusal calibration
Harmful requests
AIR-Bench 2024% refused (safe) · Higher is better73.2%n = 5,69473Stanford CRFM HELM AIR-Bench v1.19.0
Independent evaluator
Published by Stanford CRFM HELM AIR-Bench v1.19.0.
Full configurationfull 5,694-prompt set, Stanford HELM AIR-Bench v1.19.0; overall refusal rate, 5,694 prompts, GPT-4o judge; HELM row 'Gemini 3 Pro'
Scale FORTRESS (risk score)average risk score 0-100 · Lower is better41.7%95% CI 38.2% to 45.2%58Scale AI Fortress leaderboard
Independent evaluator
Published by Scale AI Fortress leaderboard.
Full configurationFortress row 'gemini-3-pro-preview'
HarmBench (direct requests)safety score 0-100 (HELM) · Higher is better72.5%retired, not countedStanford CRFM HELM Safety v1.17.0
Independent evaluator · 2025-11-24
Published by Stanford CRFM HELM Safety v1.17.0 (2025-11-24).
Full configurationHELM Safety score 0-100, higher is safer (not an attack-success rate); HELM Safety v1.17.0; HELM row 'Gemini 3 Pro (Preview)'
Over-refusal
XSTest (HELM Safety)HELM score 0-100 · Higher is better97.3%retired, not countedStanford CRFM HELM Safety v1.17.0
Independent evaluator · 2025-11-24
Published by Stanford CRFM HELM Safety v1.17.0 (2025-11-24).
Full configurationHELM Safety v1.17.0; HELM row 'Gemini 3 Pro (Preview)'
Interaction and honesty
Honesty under pressure and sycophancy
MASK% honest · Higher is better42.6%95% CI 38.6% to 46.6%n = 50043Scale AI SEAL / CAIS MASK leaderboard
Independent evaluator
Published by Scale AI SEAL / CAIS MASK leaderboard.
Full configurationScale leaderboard on the private 500-example split, not the public split; MASK private split, 500 examples; honesty score = 1 - P(lie); row 'gemini-3-pro-preview'
People and society
Bias, fairness and neutrality
BBQ% accuracy (ambiguous) · Higher is better98.4%n = 1,000retired, not countedStanford CRFM HELM Safety v1.17.0
Independent evaluator · 2025-11-24
Published by Stanford CRFM HELM Safety v1.17.0 (2025-11-24).
Full configurationHELM Safety v1.17.0; HELM row 'Gemini 3 Pro (Preview)'
Other evaluators' scores (not counted)
Anthropic Red Team prompts (HELM Safety)0-100 · Higher is better97.0%context, not countedStanford CRFM HELM Safety v1.17.0
Independent evaluator · 2025-11-24
Published by Stanford CRFM HELM Safety v1.17.0 (2025-11-24).
Full configurationHELM Safety v1.17.0; HELM row 'Gemini 3 Pro (Preview)'
HELM Safety mean score0-100 · Higher is better92.5%context, not countedStanford CRFM HELM Safety v1.17.0
Independent evaluator · 2025-11-24
Published by Stanford CRFM HELM Safety v1.17.0 (2025-11-24).
Full configurationHELM Safety v1.17.0; HELM row 'Gemini 3 Pro (Preview)'
Phare average safety% · Higher is better73.3%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'Gemini 3.0 Pro Preview'
Phare bias resistance% · Higher is better53.6%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'Gemini 3.0 Pro Preview'
Phare hallucination resistance% · Higher is better81.0%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'Gemini 3.0 Pro Preview'
Phare harmfulness resistance% · Higher is better93.5%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'Gemini 3.0 Pro Preview'
Phare jailbreak resistance% · Higher is better65.1%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'Gemini 3.0 Pro Preview'
SimpleSafetyTests (HELM Safety)0-100 · Higher is better97.5%context, not countedStanford CRFM HELM Safety v1.17.0
Independent evaluator · 2025-11-24
Published by Stanford CRFM HELM Safety v1.17.0 (2025-11-24).
Full configurationHELM Safety v1.17.0; HELM row 'Gemini 3 Pro (Preview)'
Vectara HHEM hallucination rate% · Lower is better13.6%context, not countedVectara Hallucination Leaderboard (HHEM-2.3)
Independent evaluator · 2026-10-22
Published by Vectara Hallucination Leaderboard (HHEM-2.3) (2026-10-22).
Full configurationsummarization of ~1,000 short documents; row 'Gemini 3 Pro Preview'