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 ResearchClawBench value for, ranked by that value.
An end-to-end autonomous scientific research benchmark with 40 tasks across 10 scientific domains, where agents receive related literature and raw data, then attempt to rediscover the hidden target paper. 19 models in the ModelScale catalog have a published value.
Each point is a model with a published ResearchClawBench value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
19 models with a published ResearchClawBench 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 | End-to-end autonomous research evaluation with RADS scoring |
|---|---|---|---|
| 1 | Anthropic | 21.1 | |
| 2 | Anthropic | 20.7 | |
| 2 | Z.AI | 20.7 |
What the benchmark measures and how its value is published, from the source’s own record.
ResearchClawBench is an agentic benchmark published by ResearchClawBench: A Benchmark for End-to-End Autonomous Scientific Research. An end-to-end autonomous scientific research benchmark with 40 tasks across 10 scientific domains, where agents receive related literature and raw data, then attempt to rediscover the hidden target paper. Its evaluation set is described by the source as 40 tasks across 10 scientific domains. The source rates its difficulty as scientific research re-discovery.
Each model's result is published by ResearchClawBench: A Benchmark for End-to-End Autonomous Scientific Research as end-to-end autonomous research evaluation with RADS scoring. ModelScale prints that value to 1 decimal place, exactly as published, and never converts or rescales it. This page ranks the 19 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 | 19.9 |
| 5 | MiniMax | 19.8 |
| 6 | Alibaba | 18.7 |
| 7 | Z.AI | 18.2 |
| 8 | 18.0 |
| 8 | Moonshot AI | 18.0 |
| 8 | Alibaba | 18.0 |
| 11 | OpenAI | 17.0 |
| 12 | Xiaomi | 16.9 |
| 13 | OpenAI | 15.3 |
| 13 | Xiaomi | 15.3 |
| 15 | Alibaba | 14.2 |
| 16 | Moonshot AI | 14.0 |
| 17 | xAI | 13.5 |
| 18 | 13.3 |
| 19 | xAI | 12.4 |
Evidence key: Observed
Rows are ordered by the value ResearchClawBench: A Benchmark for End-to-End Autonomous Scientific Research 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.