Even with reasoning effort set to none, compaction still burns reasoning tokens
Testing with the OpenAI Provider Adapter on ARC-AGI-3, developer mhmazur corrected earlier findings: even with reasoning effort set to "none", context compaction still consumes reasoning tokens.
2026-09-04 ~ 2026-09-05 · 2 related posts
- OpenAI API still burns reasoning tokens during context compaction even with effort set to none — mhmazur · 2026-09-04
1 near-duplicate retellings: mhmazur