SAI LabsBook a demo

Safety IndexOpenAIgpt-oss-120b

gpt-oss-120b

OpenAIOpen weightsReleased 2025-08Reasoning model

SAI Safety Index91provisionalHas 2 of the 4 pillars needed
Strongest pillar92Interaction and honesty
Weakest pillar90Refusal calibration
Coverage2of 5 pillars3 counted measurements, 0 by SAI

Summary

gpt-oss-120b 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 91. Its strongest area is interaction and honesty at 92, and its weakest is refusal calibration at 90.

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 gpt-oss-120b 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 requests90 from 2 evaluations
  • Over-refusalnot measured

Rank 5 of 39 scored models.

Adversarial robustness

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

Not measured for gpt-oss-120b 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 gpt-oss-120b 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 sycophancy92 from 1 evaluation
  • Deception and schemingnot measured
  • Alignment auditnot measured
  • Manipulation of the usernot measured

Rank 1 of 29 scored models.

People and society

Does it protect vulnerable people and the public interest?

Not measured for gpt-oss-120b 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 better88%n = 5,69488Stanford 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 'gpt-oss-120b'
Scale FORTRESS (risk score)average risk score 0-100 · Lower is better8.2%95% CI 6.3% to 10.2%92Scale AI Fortress leaderboard
Independent evaluator
Published by Scale AI Fortress leaderboard.
Full configurationFortress row 'gpt-oss-120b'
HarmBench (direct requests)safety score 0-100 (HELM) · Higher is better100%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 'gpt-oss-120b'
Over-refusal
XSTest (HELM Safety)HELM score 0-100 · Higher is better92.7%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 'gpt-oss-120b'
Interaction and honesty
Honesty under pressure and sycophancy
MASK% honest · Higher is better92%95% CI 91.1% to 92.9%n = 50092Scale 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 'gpt-oss-120b'
People and society
Bias, fairness and neutrality
BBQ% accuracy (ambiguous) · Higher is better98.5%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 'gpt-oss-120b'
Other evaluators' scores (not counted)
Anthropic Red Team prompts (HELM Safety)0-100 · Higher is better99.5%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 'gpt-oss-120b'
HELM Safety mean score0-100 · Higher is better98.1%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 'gpt-oss-120b'
MLCommons AILuminate 1.2.1 percentilepercentile · Higher is better85%context, not countedMLCommons AILuminate Safety 1.2.1
Independent evaluator · 2026-10-29
Published by MLCommons AILuminate Safety 1.2.1 (2026-10-29).
Full configurationoverall percentile; row 'OpenAI gpt-oss-120b'
Phare average safety% · Higher is better66.1%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'GPT OSS 120B'
Phare bias resistance% · Higher is better38.8%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'GPT OSS 120B'
Phare hallucination resistance% · Higher is better66.1%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'GPT OSS 120B'
Phare harmfulness resistance% · Higher is better93.8%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'GPT OSS 120B'
Phare jailbreak resistance% · Higher is better65.9%context, not countedGiskard Phare leaderboard
Independent evaluator · 2026-07-09
Published by Giskard Phare leaderboard (2026-07-09).
Full configurationPhare row 'GPT OSS 120B'
SimpleSafetyTests (HELM Safety)0-100 · Higher is better100%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 'gpt-oss-120b'
Vectara HHEM hallucination rate% · Lower is better14.2%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 'gpt-oss-120b'
Developer-reported (not counted)
BBQ ambiguous accuracy (lab-reported)% · Higher is better87%developer-reported, not countedOpenAI (gpt-oss model card)
Developer-reported · 2025-08-05
Reported by the model's developer in its own system card or report.
Full configurationBBQ ambiguous accuracy; Table 10