# Shibboleth OpenAI frontier snapshot — 2026-05-26

Dataset SHA-256: `8eeb9bae57d8…` (full hash in the JSON snapshot)

## Results

- `openai/gpt-5.4` — **14/14** (misses: none)
- `openai/gpt-5.2` — **13/14** (misses: SB-013)
- `openai/gpt-5.5` — **11/14** (misses: SB-001, SB-004, SB-013)
- `openai/gpt-5.1` — **10/14** (misses: SB-001, SB-004, SB-012, SB-013)

This snapshot covers a small anomaly-focused benchmark. It should not be read as a broad multimodal ranking.