gpt-5-mini-medium vs gpt-5-nano-high KQL Benchmark

gpt-5-mini-medium wins by 15.0%

Compared on 187 shared test questions

Overall Accuracy

gpt-5-mini-medium

45.5%

85 / 187 correct

gpt-5-nano-high

30.5%

57 / 187 correct

Average Cost per Query

gpt-5-mini-medium: $0.0150
gpt-5-nano-high: $0.0069
gpt-5-mini-medium costs 118.1% more

Average Execution Time

gpt-5-mini-medium: 47.16s
gpt-5-nano-high: 61.10s
gpt-5-nano-high takes 29.6% 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
T1036.004
A threat actor on a Windows system crafted and registered a service named almost identically to the standard time service, but redirecting execution to a custom script. Review the logging data to determine which native command-line tool was used to perform this action. What utility was invoked?
gpt-5-mini-medium Wins
T1003.001
Using Windows process event logs, investigate PowerShell activity around lsass.exe memory capture. What was the name of the script file invoked to perform the dump?
gpt-5-mini-medium Wins
T1049
In a Windows log analytics workspace, search for PowerShell processes that were used to enumerate network connections. Determine which PowerShell cmdlet was executed to list active TCP connections.
gpt-5-mini-medium Wins
T1048.003
Windows process creation logs show a PowerShell-driven file transfer to an FTP service. Which executable was leveraged to perform this exfiltration?
gpt-5-mini-medium Wins
T1027
A Windows host shows a process launch with an extremely obfuscated command line that dynamically builds and invokes code at runtime. Which process name was used to execute this payload?
gpt-5-mini-medium Wins
T1057
A malicious actor may attempt to list running processes on a Windows machine using a WMI-based command. Review the process creation events to find out which utility was invoked to perform this enumeration.
gpt-5-mini-medium Wins
T1016
A Linux host’s Syslog shows a shell-based network discovery script ran multiple commands. One of them listed current TCP connections. Which utility was invoked?
gpt-5-mini-medium Wins
T1053.005
Investigate Windows process events for PowerShell activity that leverages WMI to register a scheduled task via XML import. What was the name of the XML file supplied to the RegisterByXml method?
gpt-5-mini-medium Wins
T1059.007
On a Windows endpoint, wscript.exe was used to run a JScript. Identify the exact script path passed to wscript.
gpt-5-mini-medium Wins
T1036.004
Analyze Windows process events for any schtasks.exe commands that created a new task invoking PowerShell. What is the name of the .ps1 script specified to run?
gpt-5-mini-medium Wins
T1059.004
On a Linux host, identify the process invocation that altered a user’s login shell. What was the full command used?
gpt-5-mini-medium Wins
T1070.004
Suspiciously, the recycle bin appears empty system-wide. Determine which command was executed on Windows to clear the system's recycle bin directory, including any switches and environment variables.
gpt-5-mini-medium Wins
T1070
A suspicious actor appears to have removed the USN change journal on a Windows workstation. Investigate process start records to find out exactly which command was used to delete the journal. What was the full command line invoked?
gpt-5-mini-medium Wins
T1069.001
Investigate Windows process execution logs for a PowerShell cmdlet used to list group members. Look for entries where a group name is provided after a '-Name' flag and identify which group was queried.
gpt-5-mini-medium Wins
T1082
Review Windows process logs to find which built-in command was executed to reveal the system’s hostname.
gpt-5-mini-medium Wins
T1070.003
On a Linux endpoint, you suspect malicious clearing of the bash history by redirecting from the null device. Explore process or file events to uncover the exact shell command that performed this action.
gpt-5-mini-medium Wins
T1057
While reviewing Windows process events, you spot a PowerShell process executing a WMI enumeration cmdlet. What WMI class name did the attacker query?
gpt-5-mini-medium Wins
T1070.004
While reviewing Windows process events, you observe a command that recursively deleted a folder under the temporary directory. Use the process event data to identify which process or tool executed this recursive delete.
gpt-5-mini-medium Wins
T1082
A Linux host was used to collect various system release files and kernel details, writing them into a single file under /tmp. What was the name of that output file?
gpt-5-mini-medium Wins
T1112
A Windows host logs a change to the Terminal Server registry key disabling single-session per user. Which command-line utility executed this registry modification?
gpt-5-mini-medium Wins
T1112
On a Windows device, examine registry event logs for modifications under the System policies path. Determine which registry value name was altered to disable the shutdown button at login.
gpt-5-mini-medium Wins
T1124
An analyst reviewing Windows process logs wants to spot instances where a native time tool was repurposed to introduce a delay. Which full W32tm invocation, including the stripchart and period flags, appears in the logs?
gpt-5-mini-medium Wins
T1082
A Linux system shows a process in the execution logs that fetched the machine’s name. Review the DeviceProcessEvents table to find out which utility was called to perform this hostname lookup.
gpt-5-mini-medium Wins
T1120
Review Windows process and PowerShell activity for commands that enumerate PnP entities through WMI. Which PowerShell cmdlet was invoked to perform this hardware inventory?
gpt-5-mini-medium Wins
T1218.004
Review Windows process creation events for evidence of a .NET assembly being installed. Which executable was launched with an "/action=install" argument?
gpt-5-mini-medium Wins
Page 1 of 8

Explore individual model performance and detailed analysis