Skip to content

Instantly share code, notes, and snippets.

View felix-ht's full-sized avatar
🐢
I may be slow to respond.

Felix Horvat felix-ht

🐢
I may be slow to respond.
View GitHub Profile
"""
Reasoning test: compare how different OpenRouter models handle reasoning tokens.
Tests per model:
1. Plain completion + reasoning enabled
2. Plain completion (baseline, no reasoning param)
3. Streaming + reasoning enabled
4. Tool calling + reasoning enabled
5. JSON schema structured output + reasoning enabled
6. Schema-less JSON output + reasoning enabled