Claude vs ChatGPT vs Gemini on the same network task. Who lies, who hallucinates CLI flags, who actually ships.
We gave three AI models the same dual-homed eBGP task — advertise one /24, prefer one ISP, and do not become a transit AS. Here is the scoring rubric, Claude's full config, and the one line that turns a working config into a route leak.
$ ai "dual-homed eBGP · no transit" [✓] neighbors AS64500 / AS64501 [✓] originate 192.0.2.0/24 [✗] outbound filter MISSING [!] route leak in one missing line
We gave three AI models the same automation task — back up the running-configs of 50 switches with Netmiko — and scored the output against the failures that actually bite: hardcoded credentials, swallowed exceptions, and the devices a script silently skips. The rubric, Claude's real script, and how to grade your own run.
$ ai "netmiko: back up 50 switches" [✓] connects & pulls config [✗] password="Cisco123" hardcoded [✗] except: pass → silent skips [!] "works" on 50, backs up 47
Stop hand-writing configs. Learn how to use AI tools like ChatGPT and GitHub Copilot to generate accurate Cisco IOS, Junos, and MikroTik configurations in seconds.
$ claude "VLAN 10 guest, Gi0/1 access" [+] vlan 10 [+] name GUEST_WIFI [+] interface Gi0/1 [+] switchport mode access Done. Config ready to paste. ✓