Skip to content

Refine temperature/top_p handling for reasoning models #2277

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
mayeco wants to merge 4 commits into
base: main
Choose a base branch
from
Open

Refine temperature/top_p handling for reasoning models #2277

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
846b269
Refine temperature/top_p handling for reasoning models
mayeco Mar 3, 2026
9aab1fc
Apply suggestion from @Copilot
mayeco Mar 3, 2026
de10a5a
Update openhands-sdk/openhands/sdk/llm/options/chat_options.py
mayeco Mar 3, 2026
eb6dc3a
Merge branch 'main' into patch-3
mayeco Mar 5, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 8 additions & 3 deletions openhands-sdk/openhands/sdk/llm/options/chat_options.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,9 +39,14 @@ def select_chat_options(
if llm.reasoning_effort is not None:
out["reasoning_effort"] = llm.reasoning_effort

# All reasoning models ignore temp/top_p
out.pop("temperature", None)
out.pop("top_p", None)
# OpenAI reasoning models (o1, o3) ignore temp/top_p
# Gemini models DO respect temperature, so we only pop for OpenAI
model_lower = llm.model.lower()
# Normalize to basename so provider-prefixed IDs like "openai/o1" are handled
model_name = model_lower.split("/")[-1]
if model_name.startswith(("o1", "o3")):
Copy link
Collaborator

@enyst enyst Mar 5, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think it’s not only o1, o3, there are also Claudes, open models, GPTs

I’d actually suggest that since we removed temperature by default, we could remove this temperature special cases for all

The other models will also get nothing set (from the SDK), so they all should respect user choice

out.pop("temperature", None)
out.pop("top_p", None)

# Gemini 2.5-pro default to low if not set
if "gemini-2.5-pro" in llm.model:
Expand Down
Loading