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
Every model the catalog carries a published Terminal-Bench-Science 0.1 value for, ranked by that value.
A benchmark of AI agents completing expert-curated research workflows across the life, physical, Earth, mathematical, and engineering sciences. 15 models in the ModelChoice catalog have a published value.
Each point is a model with a published Terminal-Bench-Science 0.1 value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
15 models with a published Terminal-Bench-Science 0.1 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 | Resolution rate across 3 trials per task |
|---|---|---|---|
| 1 | OpenAI | 77.1 | |
| 2 | Anthropic | 75.3 | |
| 3 | Anthropic | 74.6 |
What the benchmark measures and how its value is published, from the source’s own record.
Terminal-Bench-Science 0.1 is an agentic benchmark published by Terminal-Bench-Science 0.1. A benchmark of AI agents completing expert-curated research workflows across the life, physical, Earth, mathematical, and engineering sciences. Its evaluation set is described by the source as 70 expert-curated scientific research workflows. The source rates its difficulty as frontier scientific research workflows.
Each model's result is published by Terminal-Bench-Science 0.1 as resolution rate across 3 trials per task. ModelChoice prints that value to 1 decimal place, exactly as published, and never converts or rescales it. This page ranks the 15 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 | OpenAI | 68.0 |
| 5 | OpenAI | 64.6 |
| 6 | Anthropic | 59.9 |
| 7 | Anthropic | 58.7 |
| 8 | OpenAI | 58.3 |
| 9 | OpenAI | 57.0 |
| 10 | Anthropic | 52.6 |
| 11 | 51.4 |
| 12 | Moonshot AI | 38.1 |
| 13 | Baidu | 21.0 |
| 14 | Tencent | 19.8 |
| 15 | TMInkling 2 Mini | Thinking Machines Lab | 1.4 |
Evidence key: Observed
Rows are ordered by the value Terminal-Bench-Science 0.1 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.