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.
Coding benchmark
Every model the catalog carries a published SWE-Atlas Codebase QnA value for, ranked by that value.
A code-comprehension benchmark covering production repositories across Go, Python, C, and TypeScript. 8 models in the ModelChoice catalog have a published value.
Each point is a model with a published SWE-Atlas Codebase QnA value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
8 models with a published SWE-Atlas Codebase QnA 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 | Mean pass@1 |
|---|---|---|---|
| 1 | InclusionAI | 64.5 | |
| 2 | Meta | 59.4 | |
| 3 | Meituan | 58.3 |
What the benchmark measures and how its value is published, from the source’s own record.
SWE-Atlas Codebase QnA is a coding benchmark published by Muse Spark 1.3 Evaluation Methodology. A code-comprehension benchmark covering production repositories across Go, Python, C, and TypeScript. Its evaluation set is described by the source as 124 codebase questions across 11 repositories. The source rates its difficulty as production codebase comprehension.
Each model's result is published by Muse Spark 1.3 Evaluation Methodology as mean pass@1. ModelChoice prints that value to 1 decimal place, exactly as published, and never converts or rescales it. This page ranks the 8 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.
| 4 | Meta | 58.0 |
| 5 | Cognition | 50.0 |
| 6 | RABeam 1.0 501B-A23B FP8 | Reflection AI | 32.2 |
| 7 | RABeam 1.0 501B-A23B | Reflection AI | 31.8 |
| 8 | RABeam 1.0 501B-A23B NVFP4 | Reflection AI | 31.4 |
Evidence key: Observed
Rows are ordered by the value Muse Spark 1.3 Evaluation Methodology 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.