Re: https://github.com/OpenHands/software-agent-sdk/pull/1497 GPT models are so many it gets confusing, I'd just like to clarify this:

"gpt-5.2-high-reasoning": {
	"id": "gpt-5.2-high-reasoning",
	"display_name": "GPT-5.2 High Reasoning",
	"llm_config": {"model": "litellm_proxy/openai/gpt-5.2-pro-2025-12-11"},

This json associates 5.2 with a 5.2-pro config, which I don't think is accurate? - GPT-5.2: - prices are at most

$1.75 / $14

-

reasoning

parameter can be

low

,

medium

,

high

,

xhigh

- can be

-codex

or not (different variants) - can be

-mini

- GPT-5.2-Pro is different than all of them. - it's their flagship - pricing is

$21 / $168

- I understand people are loving it for an architecture discussion, debugging of a complex issue, detailed planning. - like

Opus 3

was back in its day. To note,

Opus 3

pricing was kinda in this ballpark:

$15 / $75

I'd suggest maybe we can name configs that don't confuse the two. GPT-5.2-Pro is not, to the best of my knowledge, the same or close to GPT-5.2 with

reasoning='high'

. It thinks longer, and it does appear to be much better given a complex issue. "Pro" versions used to not be available via API, but 5.2 is. On this note, I think I've seen the AH litellm proxy has the name "-Pro" for some older GPTs, is it really mapped to a Pro model?