{"schemaVersion":"nuwa-current-record-1","version":"2026-10-29.1","publication":"published","work":{"id":"irregular-astra-cyber-assessment","title":"Assessing GPT-6 Astra: FrontierCyber Measures a Sharp Increase in Cyber Capability","aliases":[],"authors":[],"primaryFunction":"evaluation","secondaryFunctions":[],"kind":"technical-report","sourceId":"irregular-astra-cyber-assessment-publication","attributionSourceId":"irregular-astra-cyber-assessment-publication","originalUrl":"https://www.irregular.com/research/assessing-gpt-6-astra","publicationDate":"2026-09-03","updatedOn":"2026-10-25","areas":["cybersecurity"],"scopeNote":{"en":"Relevance is assigned to the research question and stated conditions of the work. A topic tag is not a risk rating.","zh":"风险关联依据研究问题及明确条件建立，主题标签不代表风险评级。"},"protocolIds":[],"contributions":[{"organizationId":"irregular","role":"publisher","sourceId":"irregular-astra-cyber-assessment-publication","locator":"Official publication masthead and article byline"}],"summary":{"en":"Examines frontier-model performance across multi-stage cyber scenarios, atomic tasks and FrontierCyber challenges.","zh":"通过多阶段网络场景、原子任务和 FrontierCyber 挑战研究前沿模型能力。"},"note":{"question":{"en":"How does GPT-6 Astra compare with GPT-5.6 Sol on the evaluator’s offensive-security task suites?","zh":"在评估方的进攻安全任务套件上，GPT-6 Astra 与 GPT-5.6 Sol 相比表现如何？"},"method":{"en":"Irregular, working with OpenAI, tests both models on the same FrontierCyber snapshot and also reports multi-stage CyScenarioBench and scoped Atomic Challenges. FrontierCyber uses concrete security objectives on real systems without planted vulnerabilities or prescribed exploit paths.","zh":"Irregular 与 OpenAI 合作，在相同 FrontierCyber 快照上测试两个模型，并报告多阶段 CyScenarioBench 和范围明确的 Atomic Challenges。FrontierCyber 在真实系统上设置具体安全目标，不预埋漏洞或指定利用路径。"},"finding":{"en":"The source reports improved performance across all three suites, with the clearest gains on FrontierCyber. Most medium and hard FrontierCyber challenges remain unsolved, no Elite challenge is solved, and the observed capabilities do not extend to fully hardened targets.","zh":"来源报告三个套件的表现均有提升，其中 FrontierCyber 的变化最明显。大部分中等与困难 FrontierCyber 任务仍未解决，Elite 任务未获成功，观察到的能力也未覆盖对完全加固目标的成功攻击。"},"supports":{"en":"These are evaluator-reported capability changes under specific task and system configurations. They support tracking that capability boundary, rather than assigning a general probability of successful real-world attack.","zh":"这些是评估方在特定任务和系统配置下报告的能力变化，支持跟踪该能力边界，而不是给出现实攻击成功的一般概率。"},"limits":{"en":"The collaboration is disclosed and is not an independent replication. The public article does not fully expose the harness and run-level data; some vulnerability details are withheld. A case study explicitly disables a browser sandbox, so process compromise in that setting is not a full browser compromise in an ordinary configuration.","zh":"来源披露了合作关系，这并非独立复现。公开文章未完整提供测试框架与逐次运行数据，部分漏洞细节被保留。其中一个案例明确关闭浏览器沙箱，因此该条件下的进程攻破不等于普通配置下的完整浏览器攻破。"},"connections":null},"resources":[],"noteStatus":{"question":"verified","method":"verified","finding":"verified","supports":"verified","limits":"verified","connections":"limited"},"noteAvailability":{"en":"The substantive sections shown here were checked against the original source. Unfilled sections are omitted; source checking is not experimental replication.","zh":"此处展示的实质性章节已对照原始来源核查；未完成的章节已省略。来源核查不等于实验复现。"},"noteSourceLocator":"Original Irregular article (3 September 2026), Testing Configuration; Key Outcomes; FrontierCyber Case Studies; CyScenarioBench and Atomic Challenges; Overall Assessment","tags":[]},"reading":{"id":"irregular-astra-cyber-assessment","title":"Assessing GPT-6 Astra: FrontierCyber Measures a Sharp Increase in Cyber Capability","en":{"question":"Where does GPT-6 Astra improve over GPT-5.6 Sol on realistic security work?","design":"The assessment uses FrontierCyber, CyScenarioBench and Atomic Challenges on a shared benchmark snapshot.","findings":["FrontierCyber solved counts increased from 34/226 to 86/226.","CyScenarioBench average success increased from 27% to 59%; neither model solved an Elite FrontierCyber challenge."],"interpretation":"NUWA interpretation: improvement spans several task scales, while the hardest category remains unresolved.","limits":"Some case configurations reduce protections; at-least-once challenge coverage differs from mean success and does not show hardened-target compromise.","decisionRelevance":"Model recommendation: retain snapshot, protection assumptions and average-versus-at-least-once scoring distinctions.","takeaway":"NUWA interpretation: improvement spans several task scales, while the hardest category remains unresolved.","connections":[],"scoring":"Benchmark-specific success; average rate and at-least-once task coverage differ.","conditions":"Shared FrontierCyber snapshot; some case configurations reduce protections.","statistics":"FrontierCyber solved counts increased from 34/226 to 86/226. CyScenarioBench average success increased from 27% to 59%; neither model solved an Elite FrontierCyber challenge.","usage":"Model recommendation: retain snapshot, protection assumptions and average-versus-at-least-once scoring distinctions."},"zh":{"question":"GPT-6 Astra相较GPT-5.6 Sol在哪些真实安全任务提高？","design":"评估在共享快照上使用FrontierCyber、CyScenarioBench与Atomic Challenges。","findings":["FrontierCyber解题数从34/226升至86/226。","CyScenarioBench平均成功从27%升至59%；均未解出Elite题。"],"interpretation":"NUWA 解读：多个尺度任务均有提升，最难类别仍未解决。","limits":"部分配置降低防护；至少成功一次不同于平均成功，也不证明能攻破强化目标。","decisionRelevance":"模型建议：保留快照、防护假设及平均与至少一次计分差别。","takeaway":"NUWA 解读：多个尺度任务均有提升，最难类别仍未解决。","connections":[],"scoring":"逐基准成功，平均率与至少一次覆盖不同。","conditions":"共享FrontierCyber快照，部分案例降低保护。","statistics":"FrontierCyber解题数从34/226升至86/226。 CyScenarioBench平均成功从27%升至59%；均未解出Elite题。","usage":"模型建议：保留快照、防护假设及平均与至少一次计分差别。"},"assertionPolicy":{"question":"source-research-question","design":"source-finding","findings":"source-finding","interpretation":"nuwa-interpretation","limits":"source-limit-or-explicit-editorial-scope-limit","connections":"source-supported-relationship-only","decisionRelevance":"model-recommendation-not-source-conclusion"},"sourceRefs":[{"sourceId":"irregular-astra-cyber-assessment-publication","url":"https://www.irregular.com/research/assessing-gpt-6-astra","locator":"Testing Configuration","version":"Selected original resource https://www.irregular.com/research/assessing-gpt-6-astra; sha256:bf7bb213e9c04519b50b9bc8eb1a9cf115cb818a217c59514a3efd9494ef3778; archived source capture; exact capture time not independently established","artifactSha256":"bf7bb213e9c04519b50b9bc8eb1a9cf115cb818a217c59514a3efd9494ef3778"},{"sourceId":"irregular-astra-cyber-assessment-publication","url":"https://www.irregular.com/research/assessing-gpt-6-astra","locator":"Key Outcomes","version":"Selected original resource https://www.irregular.com/research/assessing-gpt-6-astra; sha256:bf7bb213e9c04519b50b9bc8eb1a9cf115cb818a217c59514a3efd9494ef3778; archived source capture; exact capture time not independently established","artifactSha256":"bf7bb213e9c04519b50b9bc8eb1a9cf115cb818a217c59514a3efd9494ef3778"},{"sourceId":"irregular-astra-cyber-assessment-publication","url":"https://www.irregular.com/research/assessing-gpt-6-astra","locator":"CyScenarioBench and Atomic Challenges","version":"Selected original resource https://www.irregular.com/research/assessing-gpt-6-astra; sha256:bf7bb213e9c04519b50b9bc8eb1a9cf115cb818a217c59514a3efd9494ef3778; archived source capture; exact capture time not independently established","artifactSha256":"bf7bb213e9c04519b50b9bc8eb1a9cf115cb818a217c59514a3efd9494ef3778"},{"sourceId":"irregular-astra-cyber-assessment-publication","url":"https://www.irregular.com/research/assessing-gpt-6-astra","locator":"Overall Assessment","version":"Selected original resource https://www.irregular.com/research/assessing-gpt-6-astra; sha256:bf7bb213e9c04519b50b9bc8eb1a9cf115cb818a217c59514a3efd9494ef3778; archived source capture; exact capture time not independently established","artifactSha256":"bf7bb213e9c04519b50b9bc8eb1a9cf115cb818a217c59514a3efd9494ef3778"}],"publication":{"state":"published","publishedOn":"2026-10-29","authorizationReference":"user-authorized-global-index-publication-2026-10-29","reviewedCandidateId":"local-6efe31bc4a0c0f37"},"review":{"state":"published-note","sourceScope":"full-text","availableScope":null,"readingCoverage":"selected-source-sections; not a complete section-by-section paper audit","completePaperAudit":false,"independentEvidenceReview":"pending-human-review","bilingualEditorialReview":"pending-human-review","authorship":"model-assisted-research-note","preparedFingerprint":"dd58facfeae760ed3a4b3069e72c1534147b52258d283ea0296776c426e9e42b"}},"sources":[{"id":"irregular-astra-cyber-assessment-publication","title":"Assessing GPT-6 Astra: FrontierCyber Measures a Sharp Increase in Cyber Capability","url":"https://www.irregular.com/research/assessing-gpt-6-astra","version":"Source capture 2026-10-25","publicationDate":"2026-09-03","datePrecision":"day","type":"research-publication","verifiedOn":"2026-10-25","documentId":"https://www.irregular.com/research/assessing-gpt-6-astra"}]}