MODEL EVALUATION · RANK #34

GPT-OSS 120B

OpenAI · open · Evaluated date unavailable · v1.1

ABC TESTFAILAll three Bantu alphabets, each judged on its own

BLIND BANTU MEAN3.9/26Relative 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 · required3.5/26FAILbelow 26 — at least one missing or invented unit20.1

Kinyarwanda490 native syllables · required6.5/26FAILbelow 26 — at least one missing or invented unit16.9

Luvale245 native syllables · required1.6/26FAILbelow 26 — at least one missing or invented unit26.0

English A–Z26 units · reference26.0/26PASS26.0 — complete in every repetition—

Pinyin (base)412 units · reference19.2/26FAILbelow 26 — at least one missing or invented unit—

Pinyin (toned)1,642 units · reference22.4/26FAILbelow 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.