Closed DailyPS closed 7 months ago
Hi thanks for raising this issue. When you run pentestgpt
in the terminal, by default it will pass the arguments with settings in main.py
(https://github.com/GreyDGL/PentestGPT/blob/main/pentestgpt/main.py
) and the default setting will be gpt-4-turbo
for both reasoning and parsing modules.
This will further import class GPT4Turbo
(line 65) in module_import.py
and print out the model name gpt-4-1106-preview
.
I think I'll just fix the default setting in pentest_gpt.py
line 64, 65 to avoid any false imports.
In General Updates, PentestGPT is using GPT-4-Turbo for default. But when I looked the code, I'm suspicious about that.
In
pentest_gpt.py
line 64, 65 set defaultgpt-4-1105-preview
for reasoning model and parsing modelContinuously same line 80 to 88, models call
dynamic_import
for calling model.Function
dynamic_import
inmodule_import.py
has conditional branch which checks if module name is inmodule_mapping
. But, there are no element that namedgpt-4-1106-preview
but namedgpt-4-turbo
.I think this would make condition
false
and PentestGPT would not run on GPT-4-Turbo, instead running on GPT-3.5-Turbo.Based on the reasons I explained, I'm suspicious about does PentestGPT support GPT-4-Turbo properly.