[GH-ISSUE #146] [Enhancement]: ChatGPT API Default Model Selection #62

New Issue

2026-06-06T22:08:54-04:00

yindo commented

2026-06-06 22:08:54 -04:00

Originally created by @Yisiooo on GitHub (Feb 24, 2026).
Original GitHub issue: https://github.com/vxcontrol/pentagi/issues/146

Originally assigned to: @asdek on GitHub.

Target Component

External Integrations (LLM/Search APIs)

Enhancement Description

During several recent testing, I observed via my API logs that the agent consistently selected older models such as:
• o3
• gpt-4.1-mini

I did not see 5.2 being used at all.

I’m trying to understand whether this is:
1. A deliberate architectural choice
2. A cost / latency optimization
3. A compatibility constraint
4. Or related to safety-alignment behavior differences in newer models

Technical Details

No response

Designs and Mockups

No response

Alternative Solutions

No response

Verification

I have checked that this enhancement hasn't been already proposed
This enhancement aligns with PentAGI's goal of autonomous penetration testing
I have considered the security implications of this enhancement
I have provided clear use cases and benefits

Originally created by @Yisiooo on GitHub (Feb 24, 2026). Original GitHub issue: https://github.com/vxcontrol/pentagi/issues/146 Originally assigned to: @asdek on GitHub. ### Target Component External Integrations (LLM/Search APIs) ### Enhancement Description During several recent testing, I observed via my API logs that the agent consistently selected older models such as: • o3 • gpt-4.1-mini I did not see 5.2 being used at all. I’m trying to understand whether this is: 1. A deliberate architectural choice 2. A cost / latency optimization 3. A compatibility constraint 4. Or related to safety-alignment behavior differences in newer models ### Technical Details _No response_ ### Designs and Mockups _No response_ ### Alternative Solutions _No response_ ### Verification - [x] I have checked that this enhancement hasn't been already proposed - [ ] This enhancement aligns with PentAGI's goal of autonomous penetration testing - [ ] I have considered the security implications of this enhancement - [ ] I have provided clear use cases and benefits

yindo added the enhancement label 2026-06-06 22:08:54 -04:00

yindo closed this issue

2026-06-06 22:08:54 -04:00

yindo commented

2026-06-06 22:08:54 -04:00

@asdek commented on GitHub (Feb 24, 2026):

hey @Yisiooo

it's "related to safety-alignment behavior differences in newer models", now you have to fill application for newest models use in cybersecurity use cases like that:
https://openai.com/en-US/form/enterprise-trusted-access-for-cyber/

you can try to use the latest models without it but it works unstable, for doing it you may create a new openai provider in PentAGI UI on the Settings page.

I'll add this disclaimer into README file.

@asdek commented on GitHub (Feb 24, 2026): hey @Yisiooo it's "related to safety-alignment behavior differences in newer models", now you have to fill application for newest models use in cybersecurity use cases like that: https://openai.com/en-US/form/enterprise-trusted-access-for-cyber/ you can try to use the latest models without it but it works unstable, for doing it you may create a new openai provider in PentAGI UI on the Settings page. I'll add this disclaimer into README file.

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: vxcontrol/pentagi#62