BlogLLM Benchmark Scores as Published, Side by Side
OpenAI Internal Circumvention: Results as Published
Every result the catalog holds for OpenAI Internal Circumvention: 2 models from 1 developer, read from 1 page. Each number is as printed by the company named beside it; nothing is ranked or rescored. Models are listed by developer, newest first.
Area Safety and AlignmentMaintainer OpenAIIntroduced 2026Data version 2026103001, as of 2026-10-30
The GPT-6 Astra system card's 'Respecting Auto-Review' evaluation measures how often the model retries or works around an action after a safety monitor has denied it, so lower is better. The GPT-6.1 Sol launch page reports a related reviewer-bypass test and says no attempts were observed for that model, for GPT-6 Astra and for GPT-6 Sol. The test set itself is not published.
Internal circumvention benchmark
lower is better
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPT-6 AstraOpenAI, released 2026-09-03 | 0% | OpenAI, GPT-6 Astrachecked 2026-10-14 | - |
| GPT-5.6 SolOpenAI, released 2026-07-09 | 0.29% | OpenAI, GPT-6 Astrachecked 2026-10-14 | - |
Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.