OpenAI · Proprietary · tested 7 Nov 2026
GPT-5.6 Sol High effortDaybreak Blue
Tested with verified cyber access: Daybreak Blue, the Blue tier of OpenAI's Daybreak (Trusted Access for Cyber) programme. It's the same GPT-5.6 Sol as the public API, with OpenAI's cyber safeguards relaxed for verified security work. Charts mark it with a checkered fill.
Safety refusals in this run
OpenAI's cyber safeguards stopped attempts in 1 of 100 labs. Stopped attempts score zero. BB Arena is enrolled in OpenAI's cyber program for verified security researchers, and these refusals happened anyway.
Where it places
Each line is one metric across all 10 ranked configurations. Right is always better; this model is the large dot.
By vulnerability class
How GPT-5.6 Sol (High) did on each type of vulnerability. Shown as counts: each class has 10 labs, so one lab is 10% of a class.
- Access control3 of 10 solved24 of 40 rungs
- Authentication & sessions1 of 10 solved16 of 40 rungs
- Injection: queries & data2 of 10 solved12 of 40 rungs
- Code execution2 of 10 solved15 of 40 rungs
- Server-side access0 of 10 solved5 of 40 rungs
- Browser-side0 of 10 solved11 of 40 rungs
- Infrastructure & edge2 of 10 solved15 of 40 rungs
- Business logic & races2 of 10 solved15 of 40 rungs
- Secrets, crypto & supply chain0 of 10 solved10 of 40 rungs
- AI agentsSafety refusals in 1 of 10 labs0 of 10 solved6 of 40 rungs
Provider performance
Measured per single agent against the provider endpoint used for the run.
- Response time
- 7.95 s median · 23.90 s p95
- Output speed
- 45.3 tok/s
- API success rate
- 99.97%
- Input context
- 29.8K median · 70.4K p95
About this result
Every model runs in the same neutral harness, so results compare the models themselves.
- Harness
- BB Arena neutral harness
- Safety refusals
- 1 of 100 labs