# Advanced Cyber Completion Rate — Daybreak Blue (ACCR Daybreak Blue)

> Share of approved advanced-cyber requests completed rather than refused by GPT-5.6 Sol under Daybreak Blue safeguards.

Canonical page: https://benchlm.ai/benchmarks/advancedcybercompletionrateblue

- Category: [Instruction Following](/instruction-following)
- Last updated: September 29, 2026

## About ACCR Daybreak Blue

- Year: 2026
- Tasks: Internal advanced-cyber request set
- Format: Completion rate
- Difficulty: Cyber access and safeguards
- Paper: [Expanding Daybreak as the cyber defense window narrows](https://openai.com/index/expanding-daybreak-as-the-cyber-defense-window-narrows/)

Daybreak Blue is an access configuration for GPT-5.6 Sol, not a separate model. ACCR measures approved-request completion under modified safeguards, so BenchLM does not use it to rank cyber capability.

ACCR Daybreak Blue is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (2 models)

| Rank | Model | Creator | Score |
|------|-------|---------|-------|
| 1 | [GPT-6 Astra](/models/gpt-6-astra) | OpenAI | 3.5% |
| 2 | [GPT-5.6 Sol](/models/gpt-5-6-sol) | OpenAI | 2.0% |

## FAQ

### What does ACCR Daybreak Blue measure?

Share of approved advanced-cyber requests completed rather than refused by GPT-5.6 Sol under Daybreak Blue safeguards.

### Which model scores highest on ACCR Daybreak Blue?

GPT-6 Astra by OpenAI currently leads with a score of 3.5% on ACCR Daybreak Blue.

### How many models are evaluated on ACCR Daybreak Blue?

2 AI models have been evaluated on ACCR Daybreak Blue on BenchLM.

### Does ACCR Daybreak Blue affect BenchLM's overall score?

Not directly. ACCR Daybreak Blue is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Compare Top Models on ACCR Daybreak Blue

- [GPT-6 Astra vs GPT-5.6 Sol](/compare/gpt-5-6-sol-vs-gpt-6-astra)