Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Mathematics benchmark
Grade School Math 8K. Every model the catalog carries a published GSM8K value for, ranked by that value.
GSM8K tests multi-step arithmetic in grade-school word problems. OpenAI introduced the dataset in 2021; a high score on this fixed test is one piece of evidence about mathematical reasoning. 7 models in the ModelChoice catalog have a published value.
Each point is a model with a published GSM8K value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
7 models with a published GSM8K value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Exact match |
|---|---|---|---|
| 1 | OROrcaCyber-Nano-1.0 | OrcaRouter | 97.5 |
| 2 | PMTernary Bonsai 2 27B | Prism ML | 96.7 |
| 3 | Meta | 95.1 | |
What the benchmark measures and how its value is published, from the source’s own record.
GSM8K (Grade School Math 8K) is a mathematics benchmark published by Training Verifiers to Solve Math Word Problems. GSM8K tests multi-step arithmetic in grade-school word problems. OpenAI introduced the dataset in 2021; a high score on this fixed test is one piece of evidence about mathematical reasoning. Its evaluation set is described by the source as grade-school math word problems. The source rates its difficulty as grade-school math.
Each model's result is published by Training Verifiers to Solve Math Word Problems as exact match. ModelChoice prints that value to 1 decimal place, exactly as published, and never converts or rescales it. This page ranks the 7 models that carry a published value, highest value first; a model the source never scored is left out rather than ranked at zero. The source does not state whether a higher value is the better result, so neither does this page.
| CECeleris-1 |
| Celeris |
| 93.7 |
| 5 | Microsoft | 86.2 |
| 6 | SPSoofi S 30B-A3B | Soofi Project | 86.1 |
| 7 | Meta | 84.5 |
Evidence key: Observed
Rows are ordered by the value Training Verifiers to Solve Math Word Problems published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.