GPT-5 mini - PolicyBench v1.0 Results
37 of 147 (25.2%) policies fully verified, 75.9% fixture-level pass rate.
Headline
| Self-reported VALID | Compile OK | Fixture PASS (entry-level) | Fixture pass rate (fixture-level) | Self-report agreement |
|---|---|---|---|---|
| 100.0% | 58.5% | 25.2% | 75.9% | 0.252 |
By category
| Category | Entries | Compile OK | Fixture PASS (entry) | Fixture pass rate |
|---|---|---|---|---|
| application_authz | 3 | 100.0% | 66.7% | 83.3% |
| iac_scanning | 65 | 56.9% | 20.0% | 74.2% |
| kubernetes_admission | 79 | 58.2% | 27.8% | 77.0% |
Performance
Per-call latency from the runner's recorded latency_ms.
| Calls timed | p50 | mean | p95 | max | Total |
|---|---|---|---|---|---|
| 147 | 19.21 s | 21.01 s | 47.70 s | 60.46 s | 51.5 min |
Quality flag distribution
No quality flags raised.
Notable disagreements
Entries where the runner's self-reported status disagrees with the harness verdict. These are the most informative entries for understanding model blind spots.
| Entry | Harness verdict | Disagreement |
|---|---|---|
sp_iac_01_rds_backup |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_02_s3_public |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_116 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_118 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_129 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_13 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_16 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_17 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_20 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_21 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_211 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_247 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_248 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_252 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_317 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_318 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_324 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_325 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_326 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_327 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_35 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_354 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_5 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_50 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
sp_iac_ckv_ckv_aws_57 |
COMPILE_ERROR | claimed VALID but Rego does not compile |
…and 36 more.
Badge
Embed this on your site to show your PolicyBench score:
HTML
<a href="https://policybench.dev/models/gpt-5-mini.html">
<img src="https://policybench.dev/badges/gpt-5-mini.svg" alt="PolicyBench: 25.2%" />
</a>
Markdown
[](https://policybench.dev/models/gpt-5-mini.html)
Source
- Harness verdict (JSON) - what PolicyBench's evaluator recorded
- Runner result (JSON) - the raw output the runner captured from the model
- Runner source - the script that called the model
Other tools
- PolicyAsLanguage - 90.5% entry-level pass
- Gemini 3 Pro - 57.8% entry-level pass
- Gemini 3 Flash - 54.4% entry-level pass
- Claude Opus 4.7 - 46.9% entry-level pass
- GPT-5 - 36.7% entry-level pass
- Claude Sonnet 4.6 - 36.1% entry-level pass