Axion 1.0 - HLE-130 Text Only: leaderboard
Axion 1.0 author-run ReAct environment with web search, webpage visits, scholarly retrieval and Python. Best single run of three, or five for BioMysteryBench human-difficult; benchmark-approved website restrictions apply. Full HLE restricted websites and HLE-130 text-only are separate populations. Externally reported baseline scores are not pooled. HLE-130 Text Only.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Axion 1.0-122B (Axion Run: harness=ReAct; selection=best%20of%203) | 74.6 | |
| 2 | Axion 1.0-35B (Axion Run: harness=ReAct; selection=best%20of%203) | 70 | |
| 3 | Axion 1.0-9B (Axion Run: harness=ReAct; selection=best%20of%203) | 63.1 | |
| 4 | Axion 1.0-4B (Axion Run: harness=ReAct; selection=best%20of%203) | 53.1 | |
| 5 | Qwen3.6-35B-A3B (Axion Run: harness=ReAct; selection=best%20of%203) | 47.7 | |
| 6 | Qwen3.5-122B-A10B (Axion Run: harness=ReAct; selection=best%20of%203) | 40.8 |
Interactive version: theaggregate.ai/benchmark?slug=axion-1-0-hle-130-text-only · How It Works · Data refreshed daily, snapshot 2026-11-05.