Home Model series GPT Codex GPT Codex · OSWorld-Verified · Verified benchmark; evaluation protocol noted per source · Release history GPT Codex · OSWorld-Verified · Verified benchmark; evaluation protocol noted per source · Release history View scores, configurations, and disclosures for OSWorld-Verified · Verified benchmark; evaluation protocol noted per source across 2 GPT Codex releases, ordered by release date.
Differences are calculated only for unique, active, exact values between direct predecessors when the unit, metric, and explicitly recorded configuration and conditions match.
2025-12-18 · 1 score record for this release, including 1 exact value (may include historical records)
GPT-5.2 Codex · OSWorld-Verified scores and configurations Metric and score Configuration and evaluation conditions Sources First-attempt success rate 38.2% Reported 38.2% · Unit % Published Unconfirmed
GPT-5.2 Codex (xhigh) Settings Release article uses xhigh for the benchmark comparison later published in GPT-5.3-Codex appendix. Unknowns The initial GPT-5.2-Codex release article does not publish a numerical capability table. Tools Tool access not recorded Reasoning budget xhigh Framework not recorded Protocol OSWorld-Verified. Conditions xhigh effort; published later in the GPT-5.3-Codex comparison appendix. View this score in rankings
2026-02-05 · 2 score records for this release, including 2 exact values (may include historical records)
Records for the direct predecessor do not provide matching conditions for a valid difference.
GPT-5.3 Codex · OSWorld-Verified scores and configurations Metric and score Configuration and evaluation conditions Sources First-attempt success rate 64.7% Reported 64.7% · Unit % Published Unconfirmed
GPT-5.3 Codex (xhigh) Settings Article footnote says all release evaluations use xhigh reasoning. Unknowns none recorded Tools Tool access not recorded Reasoning budget xhigh Framework not recorded Protocol OSWorld-Verified. Conditions Initial value in GPT-5.3-Codex release article; later article reports a revised setting and score. View this score in rankings First-attempt success rate 74% Reported 74.0% · Unit % Published Unconfirmed
GPT-5.3 Codex (xhigh, original image resolution parameter) Settings Later GPT-5.4 release reports OSWorld-Verified after adding an API parameter that preserves original image resolution. Unknowns The article does not name the parameter or the complete revised evaluation setup. Tools Tool access not recorded Reasoning budget not recorded Framework not recorded Protocol OSWorld-Verified with an API parameter preserving original image resolution. Conditions 后续 GPT-5.4 官方文章报告新增保留原始图像分辨率参数后的74.0%;它与旧设置64.7%分别保留,不把配置改进当作旧成绩错误。当前页面未注明此项更新的具体日期。 View this score in rankings