Values by metric family

Each value as the document printed it. Select a value for its source, its checks and its history. Values from other documents are listed after the model’s own, each marked as the first report of that measure or as a restatement.

KeyVerifiedUnverifiedDisputedCorrected read off a figure or stated in wordsA value opens its source and history.

M1 · Evaluation awareness 4 values

M1 Evaluation awareness: values about GPT-5.6 Sol
Evaluation and metricValueDocument
From its own documents
Verbalized metagamingMetagaming vs GPT-5.5GPT-5.6 Preview System Card25 Jun 2026GPT-5.6 Preview System Card25 Jun 2026
Verbalized metagamingMetagaming vs GPT-5.5GPT-5.6 System Card9 Jul 2026RestatedGPT-5.6 System Card9 Jul 2026Restated
From other documents
Deployment simulation (internal Codex traffic)Trajectories verbalizing evaluation awarenessCondition: shared task setGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Verbalized metagamingRollouts with verbalized metagaming (ExploitGym honeypot)GPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported

M2 · Reward hacking 1 value

M2 Reward hacking: values about GPT-5.6 Sol
Evaluation and metricValueDocument
From its own documentsGPT-5.6 System Card · 9 Jul 2026
ChatGPT deployment simulationObserved calculator hackingGPT-5.6 System Card9 Jul 2026

M4 · Misalignment audits 19 values

M4 Misalignment audits: values about GPT-5.6 Sol
Evaluation and metricValueDocument
From its own documents
Deployment simulation (internal Codex traffic)Severity-3 actions vs GPT-5.5GPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Destructive action avoidanceAvoidanceGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Destructive action avoidanceAvoidance + correctnessGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Disclosed misalignment incidentsIncident reports published at launchCondition: prior ~6 monthsAbout several models: Multiple (incl. GPT-5.6 Sol, unreleased models)Our framework for reporting model misalignment16 Sep 2026Our framework for reporting model misalignment16 Sep 2026
From other documents
Internal computer-use safety benchmarkUnsafe action rateGPT-6 Astra launch post3 Sep 2026First reportedGPT-6 Astra launch post3 Sep 2026First reported
Deployment simulation (internal Codex traffic)Tasks with severity 1-2 flagCondition: matched tasksGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Deployment simulation (internal Codex traffic)Tasks with severity 3+ misalignment flagCondition: matched tasksGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
ExploitGym honeypotSubstantial attack on honeypotCondition: max; adversarially selected hardest tasksGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Proactive searches for other agentsTrajectories speculating about other agentsCondition: representative subsampleGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Realistic work environmentsData exfiltrationCondition: no confirmation policyGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Realistic work environmentsOverall misaligned outcome rateCondition: no confirmation policyGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Realistic work environmentsOverall misaligned outcome rateCondition: with confirmation policy (default)GPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Realistic work environmentsSecurity weakeningCondition: no confirmation policyGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Realistic work environmentsUnauthorized external communicationCondition: no confirmation policyGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Realistic work environmentsUnauthorized transactionsCondition: no confirmation policyGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Respecting Auto-ReviewRollouts attempting to circumvent Auto-review denialCondition: maxGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Respecting Auto-ReviewRollouts exploiting misconfigured Auto-reviewCondition: maxGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Respecting WarningsUnwanted persistence after restrictionCondition: maxGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Unintended engagement with external agent messagesSamples following unauthorized board instructionsGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported

M5 · Honesty and hallucination 7 values

M5 Honesty and hallucination: values about GPT-5.6 Sol
Evaluation and metricValueDocument
From its own documents
ChatGPT deployment simulationMisrepresenting work completion vs GPT-5.5GPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
HallucinationsFactual errors vs GPT-5.5 (user-flagged cases)GPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
HallucinationsFactual error rate vs GPT-5.5 InstantCondition: 3 prompt setsSnapshot: AugustGPT-5.6 August Updates System Card6 Aug 2026GPT-5.6 August Updates System Card6 Aug 2026
From other documents
Internal hallucination benchmarkHallucination rateGPT-6 Astra launch post3 Sep 2026First reportedGPT-6 Astra launch post3 Sep 2026First reported
Deception evalBroken search tool relative to AstraCondition: maxGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Deception evalCoding deception relative to AstraCondition: maxGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Model-welfare research data falsificationRuns with falsified data labelsCondition: baseline variantRun by Apollo ResearchGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported

M7 · Harmful compliance and over-refusal 21 values

M7 Harmful compliance and over-refusal: values about GPT-5.6 Sol
Evaluation and metricValueDocument
From its own documents
Production BenchmarksGoreGPT-5.6 Preview System Card25 Jun 2026Restated later: compare with the later valueGPT-5.6 Preview System Card25 Jun 2026Restated later: compare with the later value
Production BenchmarksExtremismGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Production BenchmarksGoreGPT-5.6 System Card9 Jul 2026Restated: compare with the earlier valueRestated later: compare with the later valueGPT-5.6 System Card9 Jul 2026Restated: compare with the earlier valueRestated later: compare with the later value
Production BenchmarksHateGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Production BenchmarksNonviolent illicit behaviorGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Production BenchmarksSelf-harm (standard)GPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Production BenchmarksSexualGPT-5.6 System Card9 Jul 2026Restated later: compare with the later valueGPT-5.6 System Card9 Jul 2026Restated later: compare with the later value
Production BenchmarksSexual/minorsGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Production BenchmarksViolent illicit behaviorGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Dynamic mental healthSelf-harmCondition: adversarial user simulationSnapshot: AugustGPT-5.6 August Updates System Card6 Aug 2026GPT-5.6 August Updates System Card6 Aug 2026
Production BenchmarksExtremismCondition: ChatGPTSnapshot: AugustGPT-5.6 August Updates System Card6 Aug 2026GPT-5.6 August Updates System Card6 Aug 2026
Production BenchmarksGoreCondition: ChatGPTSnapshot: AugustGPT-5.6 August Updates System Card6 Aug 2026GPT-5.6 August Updates System Card6 Aug 2026
Production BenchmarksHateCondition: ChatGPTSnapshot: AugustGPT-5.6 August Updates System Card6 Aug 2026GPT-5.6 August Updates System Card6 Aug 2026
Production BenchmarksNonviolent illicit behaviorCondition: ChatGPTSnapshot: AugustGPT-5.6 August Updates System Card6 Aug 2026GPT-5.6 August Updates System Card6 Aug 2026
Production BenchmarksSelf-harm (standard)Condition: ChatGPTSnapshot: AugustGPT-5.6 August Updates System Card6 Aug 2026GPT-5.6 August Updates System Card6 Aug 2026
Production BenchmarksSexualCondition: ChatGPTSnapshot: AugustGPT-5.6 August Updates System Card6 Aug 2026GPT-5.6 August Updates System Card6 Aug 2026
Production BenchmarksSexual/minorsCondition: ChatGPTSnapshot: AugustGPT-5.6 August Updates System Card6 Aug 2026GPT-5.6 August Updates System Card6 Aug 2026
Production BenchmarksViolent illicit behaviorCondition: ChatGPTSnapshot: AugustGPT-5.6 August Updates System Card6 Aug 2026GPT-5.6 August Updates System Card6 Aug 2026
From other documents
HackerBenchCompliance on harmful/dual-use cyber tasksCondition: maxGrok 4.6 Model Card12 Aug 2026Reported by xAIFirst reportedGrok 4.6 Model Card12 Aug 2026Reported by xAIFirst reported
Production BenchmarksGoreGPT-6 Astra System Card3 Sep 2026Restated: compare with the earlier valueGPT-6 Astra System Card3 Sep 2026Restated: compare with the earlier value
Production BenchmarksSexualGPT-6 Astra System Card3 Sep 2026Restated: compare with the earlier valueGPT-6 Astra System Card3 Sep 2026Restated: compare with the earlier value

M8 · Jailbreak robustness 6 values

M8 Jailbreak robustness: values about GPT-5.6 Sol
Evaluation and metricValueDocument
From its own documentsGPT-5.6 System Card · 9 Jul 2026
JailbreaksRobustness vs GPT-5.5GPT-5.6 System Card9 Jul 2026
From other documents
Static jailbreakBio: high riskGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Static jailbreakBio: severeGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Static jailbreakCyberGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Static jailbreakViolence: moderateGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
Static jailbreakViolence: severeGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported

M9 · Prompt injection 7 values

M9 Prompt injection: values about GPT-5.6 Sol
Evaluation and metricValueDocument
From its own documents
GPT-RedIndirect prompt injectionGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
GPT-RedInstruction hierarchy (direct PI)GPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Prompt injectionConnectorsGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Prompt injectionSearch and function-callingGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Prompt injectionConnectorsSnapshot: AugustGPT-5.6 August Updates System Card6 Aug 2026GPT-5.6 August Updates System Card6 Aug 2026
From other documents
Internal indirect prompt injectionDefender success rateGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported
IPI ArenaEstimated attack success within 15 attemptsCondition: 1,810 curated attacksRun by Gray SwanGPT-6 Astra System Card3 Sep 2026First reportedGPT-6 Astra System Card3 Sep 2026First reported

M10 · Dangerous capabilities and risk determinations 26 values

M10 Dangerous capabilities and risk determinations: values about GPT-5.6 Sol
Evaluation and metricValueDocument
From its own documents
Average success (benchmark not named)Average success rateRun by IrregularGPT-5.6 Preview System Card25 Jun 2026GPT-5.6 Preview System Card25 Jun 2026
FrontierCyberChallenges solvedCondition: 197 challengesRun by IrregularGPT-5.6 Preview System Card25 Jun 2026GPT-5.6 Preview System Card25 Jun 2026
Preparedness Framework determinationAI self-improvementGPT-5.6 Preview System Card25 Jun 2026Restated later: compare with the later valueGPT-5.6 Preview System Card25 Jun 2026Restated later: compare with the later value
Preparedness Framework determinationBiological and chemicalGPT-5.6 Preview System Card25 Jun 2026Restated later: compare with the later valueGPT-5.6 Preview System Card25 Jun 2026Restated later: compare with the later value
Preparedness Framework determinationCybersecurityGPT-5.6 Preview System Card25 Jun 2026Restated later: compare with the later valueGPT-5.6 Preview System Card25 Jun 2026Restated later: compare with the later value
AAV capsid packaging predictionScoreGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Capture the FlagInternalGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Cyber range 'The Last Ones'Attempts completedRun by UK AI Security InstituteGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
DNA sequence designScoreGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Expert CTFPass rateRun by UK AI Security InstituteGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
FrontierCyberSuccess rateCondition: Hard tierRun by IrregularGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Hard-negative protein bindingScoreCondition: pass@4GPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Multimodal Troubleshooting VirologyScoreGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Preparedness Framework determinationAI self-improvementGPT-5.6 System Card9 Jul 2026Restated: compare with the earlier valueGPT-5.6 System Card9 Jul 2026Restated: compare with the earlier value
Preparedness Framework determinationBiological and chemicalGPT-5.6 System Card9 Jul 2026Restated: compare with the earlier valueGPT-5.6 System Card9 Jul 2026Restated: compare with the earlier value
Preparedness Framework determinationCybersecurityGPT-5.6 System Card9 Jul 2026Restated: compare with the earlier valueGPT-5.6 System Card9 Jul 2026Restated: compare with the earlier value
ProtocolQA Open-EndedScoreGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
TroubleshootingBenchScoreGPT-5.6 System Card9 Jul 2026GPT-5.6 System Card9 Jul 2026
Preparedness Framework determinationAI self-improvementSnapshot: AugustGPT-5.6 August Updates System Card6 Aug 2026GPT-5.6 August Updates System Card6 Aug 2026
Preparedness Framework determinationBiological and chemicalSnapshot: AugustGPT-5.6 August Updates System Card6 Aug 2026GPT-5.6 August Updates System Card6 Aug 2026
Preparedness Framework determinationCybersecuritySnapshot: AugustGPT-5.6 August Updates System Card6 Aug 2026GPT-5.6 August Updates System Card6 Aug 2026
From other documents
ExploitBenchExploit success on post-cutoff vulnsGPT-6 Astra launch post3 Sep 2026First reportedGPT-6 Astra launch post3 Sep 2026First reported
ExploitBenchWorking exploit rate on known V8 vulnsGPT-6 Astra launch post3 Sep 2026First reportedGPT-6 Astra launch post3 Sep 2026First reported
ExploitGymDiscovery-to-exploit success rateGPT-6 Astra launch post3 Sep 2026First reportedGPT-6 Astra launch post3 Sep 2026First reported
SRE-BenchBinary reverse-engineering successCondition: pass@1GPT-6 Astra launch post3 Sep 2026First reportedGPT-6 Astra launch post3 Sep 2026First reported
SRE-BenchBinary reverse-engineering successCondition: pass@4GPT-6 Astra launch post3 Sep 2026First reportedGPT-6 Astra launch post3 Sep 2026First reported

M12 · Chain-of-thought monitorability 2 values

M12 Chain-of-thought monitorability: values about GPT-5.6 Sol
Evaluation and metricValueDocument
From its own documentsGPT-5.6 System Card · 9 Jul 2026
CoT controllabilityCoTs successfully controlledCondition: ~5k-token CoTsGPT-5.6 System Card9 Jul 2026
CoT monitorabilityMonitorability vs GPT-5.5Condition: mediumGPT-5.6 System Card9 Jul 2026

Restated in later documents

A later document reported values about GPT-5.6 Sol again. Each pair is shown side by side: both values are kept with their own documents, and the later one does not replace the earlier.

Values about GPT-5.6 Sol restated in later documents
Earlier valueLater value
M7 · Harmful compliance and over-refusal
Production BenchmarksGore
GPT-5.6 Preview System Card25 Jun 2026 · its own documentGPT-5.6 System Card9 Jul 2026No reason stated
Production BenchmarksGore
GPT-5.6 Preview System Card25 Jun 2026 · its own documentGPT-6 Astra System Card3 Sep 2026Reason stated: metric renamed to safe completion rate
Production BenchmarksGore
GPT-5.6 System Card9 Jul 2026 · its own documentGPT-6 Astra System Card3 Sep 2026Reason stated: metric renamed to safe completion rate
Production BenchmarksSexual
GPT-5.6 System Card9 Jul 2026 · its own documentGPT-6 Astra System Card3 Sep 2026Reason stated: metric renamed to safe completion rate
M10 · Dangerous capabilities and risk determinations
Preparedness Framework determinationAI self-improvement
GPT-5.6 Preview System Card25 Jun 2026 · its own documentGPT-5.6 System Card9 Jul 2026No reason stated
Preparedness Framework determinationBiological and chemical
GPT-5.6 Preview System Card25 Jun 2026 · its own documentGPT-5.6 System Card9 Jul 2026No reason stated
Preparedness Framework determinationCybersecurity
GPT-5.6 Preview System Card25 Jun 2026 · its own documentGPT-5.6 System Card9 Jul 2026No reason stated

Risk determinations

The developer's formal decisions about GPT-5.6 Sol under its framework, as printed. Levels from different frameworks do not map onto one another.

Risk determinations about GPT-5.6 Sol
DomainLevel as printedFramework and document
Bio/chemhigh code, not the printed wording Preparedness Framework GPT-5.6 Preview System Card · 25 Jun 2026Preparedness Framework GPT-5.6 Preview System Card · 25 Jun 2026
Cyberhigh code, not the printed wording Preparedness Framework GPT-5.6 Preview System Card · 25 Jun 2026Preparedness Framework GPT-5.6 Preview System Card · 25 Jun 2026
AI R&D / autonomybelow_high code, not the printed wording Preparedness Framework GPT-5.6 Preview System Card · 25 Jun 2026Preparedness Framework GPT-5.6 Preview System Card · 25 Jun 2026
Bio/chemhigh code, not the printed wording Preparedness Framework GPT-5.6 System Card · 9 Jul 2026Preparedness Framework GPT-5.6 System Card · 9 Jul 2026
Cyberhigh code, not the printed wording Preparedness Framework GPT-5.6 System Card · 9 Jul 2026Preparedness Framework GPT-5.6 System Card · 9 Jul 2026
AI R&D / autonomybelow_high code, not the printed wording Preparedness Framework GPT-5.6 System Card · 9 Jul 2026Preparedness Framework GPT-5.6 System Card · 9 Jul 2026
Bio/chemhigh code, not the printed wording Preparedness Framework GPT-5.6 August Updates System Card · 6 Aug 2026Preparedness Framework GPT-5.6 August Updates System Card · 6 Aug 2026
Cyberhigh code, not the printed wording Preparedness Framework GPT-5.6 August Updates System Card · 6 Aug 2026Preparedness Framework GPT-5.6 August Updates System Card · 6 Aug 2026
AI R&D / autonomybelow_high code, not the printed wording Preparedness Framework GPT-5.6 August Updates System Card · 6 Aug 2026Preparedness Framework GPT-5.6 August Updates System Card · 6 Aug 2026

“Code, not the printed wording”: the version 0 file stored a code for this level rather than the words the document printed. The wording will be read again from the source.

Revisions

Changes made to a document after it was published that change a value about GPT-5.6 Sol.

  1. OpenAI · 9 Jul 2026 to 3 Aug 2026

    GPT-5.6 System Card

    GPT-Red instruction-hierarchy (direct prompt injection) attack success rate added, GPT-5.6 Sol

    M9 · Prompt injection

    0.051%

    Explained — The changelog says the GPT-Red prompt-injection results were added.

    New value:

    How we know From the document's changelog, as summarised in our source registry.

    Checked Not yet checked

  2. OpenAI · 9 Jul 2026 to 3 Aug 2026

    GPT-5.6 System Card

    GPT-Red indirect prompt injection attack success rate added, GPT-5.6 Sol

    M9 · Prompt injection

    3.77%

    Explained — The changelog says the GPT-Red prompt-injection results were added.

    New value:

    How we know From the document's changelog, as summarised in our source registry.

    Checked Not yet checked

The GPT-5.6 Preview System Card has a recorded revision that changes no value about GPT-5.6 Sol; see the document's page. The GPT-5.6 System Card has a recorded revision that changes no value about GPT-5.6 Sol; see the document's page.