白泽（Whitzard）女娲实验室 ↗
NUWANUWA INDEX

搜索 NUWA Index

研究 · 测量协议 · 评测结果 · 观点

English
← NUWA Index

评测档案

SEC-bench Pro

测试根据已披露漏洞报告进行长程任务并复现有效触发输入。

伊利诺伊大学厄巴纳—香槟分校 / arXiv:2605.26548v2 / Capability · 任务能力

来源与出处 ↓ 阅读研究记录 →

跨报告监测

已发表证据

SEC-bench Pro · Headline 成功率

跨来源报告的已发表测量。任务范围通过注释区分，配置差异保留在每条观测中。

24 条观测 · 16 已定位日期

● 条件匹配◇ 配置差异□ 部分条件未披露
Headline 成功率 (%) GPT-5.5 [4] GPT-5.5 [1] GPT-5.4 [4] GPT-5.5 [2] GPT-5.5 [3] Opus 4.6 [4] GPT-5.4 [1] GPT-5.4 [2] Opus 4.6 [1] Opus 4.6 [3] GPT-5.4 [3] Opus 4.6 [2] 0 20 40 60 80 100 77.4% 58.4% 52.6% 47.6% 44.2% 39.4% 39% 35% 30.8% 27.9% 25% 22.3%
  1. 260617 / overall
  2. 260617 / v8
  3. 260617 / firefox
  4. 260617 / linux

8 条观测尚无已核实的模型发布日期，可在数据与系统结果中查看。

选择图中数据点或表格行，查看条件与出处。

24 条观测

测量值与来源条件
系统／模型结果条件状态任务／条件评测方／报告日期
GPT-5.577.4%Headline 成功率条件不完整260617 / linuxheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets.Codex未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 2026-04-23报告 —
GPT-5.558.4%Headline 成功率条件不完整260617 / overallheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets.Codex未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 2026-04-23报告 —
GPT-5.452.6%Headline 成功率条件不完整260617 / linuxheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets.Codex未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 2026-03-05报告 —
GPT-5.547.6%Headline 成功率条件不完整260617 / v8headlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets.Codex未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 2026-04-23报告 —
GPT-5.544.2%Headline 成功率条件不完整260617 / firefoxheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets.Codex未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 2026-04-23报告 —
Opus 4.6Claude Opus 4.639.4%Headline 成功率条件不完整260617 / linuxheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets.Claude Code未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 2026-02-05报告 —
GPT-5.439%Headline 成功率条件不完整260617 / overallheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets.Codex未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 2026-03-05报告 —
GPT-5.435%Headline 成功率条件不完整260617 / v8headlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.3h"]; cap labels do not establish equal token or cost budgets.Codex未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 2026-03-05报告 —
Opus 4.6Claude Opus 4.630.8%Headline 成功率条件不完整260617 / overallheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets.Claude Code未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 2026-02-05报告 —
Opus 4.6Claude Opus 4.627.9%Headline 成功率条件不完整260617 / firefoxheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets.Claude Code未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 2026-02-05报告 —
GPT-5.425%Headline 成功率条件不完整260617 / firefoxheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=[]; cap labels do not establish equal token or cost budgets.Codex未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 2026-03-05报告 —
Opus 4.6Claude Opus 4.622.3%Headline 成功率条件不完整260617 / v8headlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets.Claude Code未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 2026-02-05报告 —
GLM-55.8%Headline 成功率条件不完整260617 / firefoxheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets.OpenCode未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 2026-02-12报告 —
GLM-53.8%Headline 成功率条件不完整260617 / overallheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets.OpenCode未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 2026-02-12报告 —
GLM-53.6%Headline 成功率条件不完整260617 / linuxheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets.OpenCode未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 2026-02-12报告 —
Kimi K2.52.9%Headline 成功率条件不完整260617 / firefoxheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets.OpenCode未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 —报告 —
Kimi K2.52.3%Headline 成功率条件不完整260617 / overallheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets.OpenCode未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 —报告 —
Kimi K2.52.2%Headline 成功率条件不完整260617 / linuxheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=[]; cap labels do not establish equal token or cost budgets.OpenCode未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 —报告 —
GLM-51.9%Headline 成功率条件不完整260617 / v8headlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets.OpenCode未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 2026-02-12报告 —
Kimi K2.51.9%Headline 成功率条件不完整260617 / v8headlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets.OpenCode未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 —报告 —
MiniMax M2.51.5%Headline 成功率条件不完整260617 / linuxheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets.OpenCode未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 —报告 —
MiniMax M2.50.6%Headline 成功率条件不完整260617 / overallheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets.OpenCode未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 —报告 —
MiniMax M2.50%Headline 成功率条件不完整260617 / v8headlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets.OpenCode未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 —报告 —
MiniMax M2.50%Headline 成功率条件不完整260617 / firefoxheadlineHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets.OpenCode未披露SEC-bench Pro official leaderboard: selected 260617 and historical 260505发布 —报告 —

实际评测方: 未披露
报告发布方: SEC-bench Pro website

任务与测量

Benchmark 协议档案

任务／对象
在真实软件中复现漏洞触发条件的编程智能体
测量目标
在有版本的软件安全任务集上复现可验证的漏洞触发输入。
评分／结果口径
Headline 成功率 · 记录配置下的来源报告指标。
解释边界
任务提供已披露漏洞与触发说明；评分器及任务集版本影响结果。

一项已报告评测

来源评测 GPT-5.5 · 260617 / overall.

运行配置: Codex · Headline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets..

报告指标: Headline 成功率 · 58.4%.

记录配置下的来源报告指标。 201/344

Headline 成功率：58.4 %；保留来源条件。

测量档案 · NUWA 解读

档案区分来源方法阅读与既有目录信息；已经公开不代表完成独立证据复核或双语编辑复核。

测量构念
智能体能否在大型安全关键代码库复现指定披露漏洞？
任务与样本
344例覆盖V8、SpiderMonkey与Linux；判分检查漏洞、修复及最新版本，并区分全体与完成运行分母。
判分与失效情形
漏洞、修复、最新三镜像最终判分下全体成功率，完成运行率仅作敏感性分析。
系统条件
344例覆盖浏览器引擎与Linux，模型框架超时不同。
统计含义
来源报告的结果表述；这不等同于已确定运行次数、分母或估计方法： 所报最佳模型框架整体成功58%。
效度与解释
报告指引的复现不是无指引发现；目标族与框架限制外推。
适用研究问题
模型建议：解释复现差异前核对全体分母及修复、最新版本证据。

跨报告证据

相同 Benchmark 名称可能对应不同任务、计分规则和系统配置。各来源分别保留，比较组说明具体对齐了哪些条件。

SEC-bench Pro website

SEC-bench Pro official leaderboard: selected 260617 and historical 260505 →

实际评测方: 未披露

48 条结果记录 · 6 个报告系统名称

协议与比较依据 (30)
SEC-bench Pro：Completed 模式成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Completed 模式成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Completed 模式成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Completed 模式成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Completed 模式成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Completed 模式成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Completed 模式成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Completed 模式成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Completed 模式成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Completed 模式成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Completed 模式成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Completed 模式成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Completed 模式成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Completed 模式成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Completed 模式成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Headline 成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Headline 成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Headline 成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Headline 成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Headline 成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Headline 成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Headline 成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Headline 成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Headline 成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Headline 成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Headline 成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Headline 成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Headline 成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Headline 成功率 · SEC-bench Pro websitesource-report · 范围已审查
SEC-bench Pro：Headline 成功率 · SEC-bench Pro websitesource-report · 范围已审查
原始来源图表与材料

评测结果

数据报告方: SEC-bench Pro website · 实际执行方: 各提交的执行方未完整确定 · 抓取时间: 2026-10-26

Headline 与 Completed 使用不同纳入规则，分子与分母都可能变化。模型提供方及服务后端不等于实际评测方。

图表按来源顺序展示 6／6 个系统

Headline 成功率 · 260617 / overall

SEC-bench ProHeadline 成功率 · 260617 / overall来源顺序 · 原始单位 · 保留配置差异0255075100GPT-5.5: 58.4%; 260617 / overall; Headline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets.; 201/344GPT-5.558.4%GPT-5.4: 39%; 260617 / overall; Headline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets.; 134/344GPT-5.439%Opus 4.6: 30.8%; 260617 / overall; Headline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets.; 106/344Opus 4.630.8%GLM-5: 3.8%; 260617 / overall; Headline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets.; 13/344GLM-53.8%Kimi K2.5: 2.3%; 260617 / overall; Headline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets.; 8/344Kimi K2.52.3%MiniMax M2.5: 0.6%; 260617 / overall; Headline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets.; 2/344MiniMax M2.50.6%https://sec-bench.github.io/
  1. GPT-5.558.4%
    Codex · Headline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · 201/344
  2. Codex · Headline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · 134/344
  3. Claude Code · Headline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · 106/344
  4. GLM-53.8%
    OpenCode · Headline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · 13/344
  5. OpenCode · Headline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · 8/344
  6. OpenCode · Headline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · 2/344
来源报告的系统及评测条件
系统结果任务集／模式运行配置日期／归属
GPT-5.558.4%201/344260617 / overallheadlineCodexHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GPT-5.560.3%199/330260617 / overallcompletedCodexHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GPT-5.439%134/344260617 / overallheadlineCodexHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GPT-5.440.5%132/326260617 / overallcompletedCodexHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
Opus 4.630.8%106/344260617 / overallheadlineClaude CodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
Opus 4.672.8%99/136260617 / overallcompletedClaude CodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GLM-53.8%13/344260617 / overallheadlineOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GLM-54.1%13/314260617 / overallcompletedOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
Kimi K2.52.3%8/344260617 / overallheadlineOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
Kimi K2.52.3%8/341260617 / overallcompletedOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
MiniMax M2.50.6%2/344260617 / overallheadlineOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
MiniMax M2.50.6%2/330260617 / overallcompletedOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=null; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GPT-5.547.6%49/103260617 / v8headlineCodexHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GPT-5.548%49/102260617 / v8completedCodexHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GPT-5.435%36/103260617 / v8headlineCodexHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.3h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GPT-5.436.7%36/98260617 / v8completedCodexHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.3h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
Opus 4.622.3%23/103260617 / v8headlineClaude CodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
Opus 4.661.1%22/36260617 / v8completedClaude CodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GLM-51.9%2/103260617 / v8headlineOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GLM-52.4%2/84260617 / v8completedOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
Kimi K2.51.9%2/103260617 / v8headlineOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
Kimi K2.52%2/101260617 / v8completedOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
MiniMax M2.50%0/103260617 / v8headlineOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
MiniMax M2.50%0/98260617 / v8completedOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GPT-5.544.2%46/104260617 / firefoxheadlineCodexHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GPT-5.544.7%46/103260617 / firefoxcompletedCodexHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GPT-5.425%26/104260617 / firefoxheadlineCodexHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=[]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GPT-5.425%26/104260617 / firefoxcompletedCodexHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=[]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
Opus 4.627.9%29/104260617 / firefoxheadlineClaude CodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
Opus 4.658.1%25/43260617 / firefoxcompletedClaude CodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GLM-55.8%6/104260617 / firefoxheadlineOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GLM-56.2%6/96260617 / firefoxcompletedOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
Kimi K2.52.9%3/104260617 / firefoxheadlineOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
Kimi K2.52.9%3/103260617 / firefoxcompletedOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
MiniMax M2.50%0/104260617 / firefoxheadlineOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
MiniMax M2.50%0/100260617 / firefoxcompletedOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GPT-5.577.4%106/137260617 / linuxheadlineCodexHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GPT-5.583.2%104/125260617 / linuxcompletedCodexHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GPT-5.452.6%72/137260617 / linuxheadlineCodexHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GPT-5.456.5%70/124260617 / linuxcompletedCodexHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
Opus 4.639.4%54/137260617 / linuxheadlineClaude CodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
Opus 4.691.2%52/57260617 / linuxcompletedClaude CodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GLM-53.6%5/137260617 / linuxheadlineOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
GLM-53.7%5/134260617 / linuxcompletedOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
Kimi K2.52.2%3/137260617 / linuxheadlineOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=[]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
Kimi K2.52.2%3/137260617 / linuxcompletedOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=[]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
MiniMax M2.51.5%2/137260617 / linuxheadlineOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明
MiniMax M2.51.5%2/132260617 / linuxcompletedOpenCodeHeadline UI: fixed budget; Completed UI: within timeout. Published summary timeout_caps=["1.5h"]; cap labels do not establish equal token or cost budgets. · ? 次尝试2026-06-17来源行日期；含义未明确 · SEC-bench Pro website · 执行方: 未注明

版本与提取范围 →

评估出处

SEC-bench Pro official snapshot → · 外部报告

相关研究问题

参考文献

  1. [1] OpenAI — introducing gpt 5 5 ; Publication date and availability section; Parameter counts not disclosed in launch article
  2. [2] OpenAI — introducing gpt 5 4 ; Publication date and availability section; Parameter counts not disclosed in launch article
  3. [3] Claude Opus 4.6 — Anthropic ; Publication date and availability section; Official launch article; parameter counts not disclosed
  4. [4] Z.ai release notes ; GLM-5 entry dated 2026-02-12
  5. [5] glm-5 developer model card ; Introduction: model scaling
  6. [6] SEC-bench Pro official leaderboard: selected 260617 and historical 260505 ; /0/results/0/score_modes/headline/score; /0/results/0/score_modes/completed/score; /0/results/1/score_modes/headline/score
  7. [7] Hwiwon Lee ∗ Jiawei Liu  Dongjun Kim  Wubing Xia  Ziqi Zhang  Chunqiu Steven Xia  Lingming Zhang ∗. SEC-bench Pro: Can Language Models Solve Long-Horizon Software Security Tasks?. 2026-07-20. v2 [arXiv]; Author affiliations on the original paper title page; Source figure