gpt-5-mini-high vs gpt-5-mini-medium KQL Benchmark
gpt-5-mini-high wins by 2.7%
Compared on 187 shared test questions
Overall Accuracy
gpt-5-mini-high
48.1%
90 / 187 correct
gpt-5-mini-medium
45.5%
85 / 187 correct
Average Cost per Query
gpt-5-mini-high: $0.0150
gpt-5-mini-medium: $0.0150
gpt-5-mini-medium costs 0.3% more
Average Execution Time
gpt-5-mini-high: 44.83s
gpt-5-mini-medium: 47.16s
gpt-5-mini-medium takes 5.2% longer
Question-by-Question Analysis
Question-by-Question Comparison
Detailed comparison showing where each model succeeded or failed
Showing 1 to 25 of 187 questions
Page 1 of 8
T1007 | An analyst suspects a user or script ran a service enumeration command on a Linux system. Review process events to find the service-listing invocation and specify the full command that was executed. | ✓ | ✗ | gpt-5-mini-high Wins |
T1018 | On a Windows endpoint, review process creation logs to uncover when a built-in utility was used to reveal ARP entries. What exact command was used to list the ARP cache? | ✓ | ✗ | gpt-5-mini-high Wins |
T1021.006 | On Windows hosts, look through recent PowerShell execution records to find any elevated session where remote management was turned on. What exact command was run to enable PSRemoting? | ✓ | ✗ | gpt-5-mini-high Wins |
T1006 | Identify the PowerShell cmdlet used on Windows to format and display the raw volume bytes after an attacker read the boot sector via a DOS device path. | ✓ | ✗ | gpt-5-mini-high Wins |
T1053.003 | Linux hosts may log events when new files are added to /var/spool/cron/crontabs. Query those logs for a creation or write action in that directory and determine the file name that was added. | ✓ | ✗ | gpt-5-mini-high Wins |
T1053.006 | Examine the logs from the Linux system for events related to the systemd timer activation. Identify any records indicating that a new timer unit was started and enabled, and determine which timer name was used. | ✓ | ✗ | gpt-5-mini-high Wins |
T1059.004 | During a Linux investigation, you notice processes spawning curl and wget commands that pull a script from a remote GitHub raw URL and pipe it into bash. Identify the name of the script that was retrieved and executed. | ✓ | ✗ | gpt-5-mini-high Wins |
T1057 | A Windows endpoint recorded a command-line activity through cmd.exe that lists all running processes. Determine which built-in tool was executed to perform this action. | ✓ | ✗ | gpt-5-mini-high Wins |
T1069.001 | On a Linux endpoint, process events reveal a chain of group‐enumeration utilities executed by a single session. Which utility was used to query the system’s group database? | ✓ | ✗ | gpt-5-mini-high Wins |
T1070.003 | On a Linux system, you suspect someone erased their command history by linking the history file to /dev/null. Investigate process events and determine which utility was executed to achieve this. | ✓ | ✗ | gpt-5-mini-high Wins |
T1070.003 | On a Windows endpoint, review process execution logs to see if any PowerShell sessions were wiped clean. Which command was executed to clear the PowerShell history? | ✓ | ✗ | gpt-5-mini-high Wins |
T1070.003 | On a Windows device, there’s evidence that PowerShell history was wiped by deleting the history file. What was the exact command used to perform this action? | ✓ | ✗ | gpt-5-mini-high Wins |
T1059.004 | An attacker on a Linux host may try to enumerate installed shells by reading the system file that lists valid shells. Using process or syslog data, determine which command was executed to perform this enumeration. | ✓ | ✗ | gpt-5-mini-high Wins |
T1070.006 | On a Windows host, suspicious PowerShell activity adjusted the system clock and recorded a value. What numeric value was used to slip the system date? | ✓ | ✗ | gpt-5-mini-high Wins |
T1078.003 | Review the Linux process creation records to find which user account management utility was used to reactivate the previously locked and expired account. | ✓ | ✗ | gpt-5-mini-high Wins |
T1082 | A Windows system shows a cmd.exe process spawn that appears to have been used for environment discovery. Review the process creation records to identify the exact command the adversary ran to enumerate environment variables. | ✓ | ✗ | gpt-5-mini-high Wins |
T1112 | Review registry event logs on the Windows host for PowerShell-driven writes to system policy and file system keys. Which registry value names were created during this BlackByte preparation simulation? | ✓ | ✗ | gpt-5-mini-high Wins |
T1112 | A Windows user’s registry was altered via a command-line tool to disable the lock workstation feature by adding a DWORD entry under the current user Policies\System key. Which registry value name was modified in this operation? | ✓ | ✗ | gpt-5-mini-high Wins |
T1112 | Investigate Windows registry events to identify any newly set ProxyServer entry under the user Internet Settings hive. What proxy server address was configured? | ✓ | ✗ | gpt-5-mini-high Wins |
T1124 | A Windows host recorded a process that simply executes the system’s native time utility. Without spelling out the query, determine which command was run based on process creation events. | ✓ | ✗ | gpt-5-mini-high Wins |
T1112 | On Windows systems, disabling RDP via the registry generates registry write events. Investigate registry event logs for modifications under the Terminal Server configuration path. What is the name of the registry value that was changed to disable Remote Desktop Protocol? | ✓ | ✗ | gpt-5-mini-high Wins |
T1134.001 | A Windows host logs show PowerShell fetching and executing a remote script to gain SeDebugPrivilege token duplication. Which Empire module was invoked? | ✓ | ✗ | gpt-5-mini-high Wins |
T1217 | An attacker leveraged a PowerShell command on a Windows host to enumerate browser bookmark files across all user profiles. Examine the process execution logs to determine the exact filename that was being searched for. | ✓ | ✗ | gpt-5-mini-high Wins |
T1490 | An attacker obtained elevated rights on a Windows system and ran a deletion command that attempted to remove various backup file types across the C: drive, generating numerous “access denied” errors. What was the full command line used? | ✓ | ✗ | gpt-5-mini-high Wins |
T1218.011 | You notice rundll32.exe being used with desk.cpl,InstallScreenSaver on a Windows endpoint. Investigate your process creation logs to find which .scr file was loaded by this unusual invocation. | ✓ | ✗ | gpt-5-mini-high Wins |
Page 1 of 8
Explore individual model performance and detailed analysis