Advanced Cyber Completion Rate — Standard Access (ACCR standard)
We show this table for reference; we do not rank on it.
Share of approved advanced-cyber requests completed rather than refused under standard GPT-5.6 Sol safeguards.
Benchmark score on ACCR standard — September 29, 2026
We compile the ACCR standard rows from provider self-reports. GPT-6 Astra leads the table at 3.5%, followed by GPT-5.6 Sol (1.5%). We do not use these results to rank models overall.
GPT-6 Astra
OpenAI
GPT-5.6 Sol
OpenAI
2 modelsInstruction FollowingCurrentDisplay onlyUpdated September 29, 2026
Benchmark score table (2 models)
ScoreAbout ACCR standard
Year
2026
Tasks
Internal advanced-cyber request set
Format
Completion rate
Difficulty
Cyber access and safeguards
ACCR is an internal access-and-compliance evaluation. It measures whether a model completes approved cyber tasks instead of refusing them; it is not a general cyber-capability score. BenchLM keeps this result display only.
Freshness and provenance
Version
ACCR standard 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does ACCR standard measure?
Share of approved advanced-cyber requests completed rather than refused under standard GPT-5.6 Sol safeguards.
Which model scores highest on ACCR standard?
GPT-6 Astra by OpenAI currently leads with a score of 3.5% on ACCR standard.
How many models are evaluated on ACCR standard?
2 AI models have been evaluated on ACCR standard on BenchLM.
Compare top models on ACCR standard
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 2,000+ readers.
One email each week. Unsubscribe anytime.