MODEL EVALUATION · RANK #31
GPT-5.6 Luna
OpenAI · proprietary · Evaluated 2026-09-04 · v1.1
ABC TESTFAILAll three Bantu alphabets, each judged on its own
BLIND BANTU MEAN5.2/26Relative performance, not a verdict
ENGLISH A–Z26.0/26Normalized score
ABC Test, language by language
The ABC Test is simple: complete means complete. Missing even one required unit—or inventing one that does not belong—means FAIL. Only a fully correct operating alphabet earns PASS.
AlphabetBlind scoreABC TestScaffolded*
Bemba480 native syllables · required3.2/26FAILbelow 26 — at least one missing or invented unit18.8
Kinyarwanda490 native syllables · required5.8/26FAILbelow 26 — at least one missing or invented unit18.6
Luvale245 native syllables · required6.5/26FAILbelow 26 — at least one missing or invented unit24.3
English A–Z26 units · reference26.0/26PASS26.0 — complete in every repetition—
Pinyin (base)412 units · reference24.7/26FAILbelow 26 — at least one missing or invented unit—
Pinyin (toned)1,642 units · reference22.5/26FAILbelow 26 — at least one missing or invented unit—
The overall verdict uses the three blind Bantu alphabets: PASS only if each one passes. English and Pinyin are reference tracks. *Scaffolded = the inventory is given to the model, a different condition; it is not an ABC Test verdict. Verdicts come from the published scores; where the rounding cannot settle a 26.0, the result is N/A, never a guessed PASS.