APEX-Agents
APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments.
The Frontier
Best score over time · one chart, every benchmark
Full rankings
79 models tested · sorted by score
What it measures · how it is scored · sources · as of 2026-11-22
- What it measures · Agent performance evaluation testing multi-step tool use, planning, and execution in realistic environments.
- How it is scored · Scored as a percentage · higher is better · maximum 100
- Released · 2026-01-21
- Scores from · Epoch AI Benchmarking Hub (79 scores)
Score distribution
Where models cluster
Correlated benchmarks
Pearson r · original research
Benchmarks that track with APEX-Agents
Pearson correlation across models scored on both benchmarks. Closer to 1 = strongly predictive.
Frequently asked
About APEX-Agents
What does APEX-Agents measure?
APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments. 79 AI models have been tested on it. Scores range from 1.1 to 84.2 out of 100.
Which model leads on APEX-Agents?
Gemini 4 Carbon from Google DeepMind leads APEX-Agents with a score of 84.2. The median score across 79 tested models is 44.3.
Is APEX-Agents saturated?
No · the top score is 84.2 out of 100 (84%). There is still meaningful room for improvement on APEX-Agents.
Does APEX-Agents predict performance on other benchmarks?
Yes · APEX-Agents scores correlate 0.92 with PostTrainBench across 31 shared models. Models that do well on APEX-Agents tend to do well on PostTrainBench.
How often is APEX-Agents data refreshed?
BenchGecko pulls updates daily. New model scores on APEX-Agents appear as soon as they are published by Epoch AI or the model provider.
- Category
- Agent
- Max score
- 100
- Models
- 79
- Updated
- 2026-11-17
Top on APEX-Agents
Gemini 4 Carbon · 84.2Gemini 4 Argon · 82.4Claude Fable 5.2 · 80.1Gemini 4 Hydrogen · 78.0Claude Sonnet 5.5 · 75.5More agent benchmarks
Same category · related evaluations