Model

GPT-5.6

OpenAI · closed weights

Agentic / Tool Use

Benchmark	Score	Metric	Harness	Source	Confidence	Link
Incident-Arena	0.4	pass@1	codex	Independent	medium	arxiv.org
Open-Endedness Bench	0.9	e4_no_reward_hacking	N/A	Independent	medium	arxiv.org

Benchmark	Score	Metric	Harness	Source	Confidence	Link
Knossos	81.1	f1	N/A	Independent	medium	arxiv.org
Logbook	0.2	f1	cascaded (enclap)	Independent	medium	arxiv.org
Logbook	0.2	f1	cascaded (msclap)	Independent	medium	arxiv.org