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.
Agentic benchmark
ApprenticeBench: end-to-end computer use, continual learning, and long-horizon agency on a real accounts-payable job. Every model the catalog carries a published ApprenticeBench value for, ranked by that value.
Tests whether a computer-use agent can learn a real accounts-payable job on the job, processing 100 vendor bills in a company ERP system with only the handbook, historical records, and mentor feedback a new hire would get. 18 models in the ModelScale catalog have a published value.
Each point is a model with a published ApprenticeBench value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
18 models with a published ApprenticeBench 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 | Cumulative success rate over 100 bills |
|---|---|---|---|
| 1 | Anthropic | 72 | |
| 2 | OpenAI | 68 | |
| 3 | Anthropic | 36 |
What the benchmark measures and how its value is published, from the source’s own record.
ApprenticeBench (ApprenticeBench: end-to-end computer use, continual learning, and long-horizon agency on a real accounts-payable job) is an agentic benchmark published by ApprenticeBench: a step change in AI's job readiness. Tests whether a computer-use agent can learn a real accounts-payable job on the job, processing 100 vendor bills in a company ERP system with only the handbook, historical records, and mentor feedback a new hire would get. Its evaluation set is described by the source as 100 vendor bills processed in sequence inside a simulated construction company. The source rates its difficulty as long-horizon computer use with offline and online continual learning.
Each model's result is published by ApprenticeBench: a step change in AI's job readiness as cumulative success rate over 100 bills. ModelScale prints that value to 0 decimal places, exactly as published, and never converts or rescales it. This page ranks the 18 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 | Anthropic | 34 |
| 5 | OpenAI | 26 |
| 6 | 24 |
| 7 | OpenAI | 20 |
| 8 | Meta | 19 |
| 9 | Moonshot AI | 18 |
| 10 | Anthropic | 16 |
| 10 | 16 |
| 10 | OpenAI | 16 |
| 13 | xAI | 13 |
| 14 | OpenAI | 11 |
| 15 | Anthropic | 7 |
| 15 | OpenAI | 7 |
| 17 | Anthropic | 5 |
| 18 | Anthropic | 2 |
Evidence key: Observed
Rows are ordered by the value ApprenticeBench: a step change in AI's job readiness 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.