CAIA - Pass@1 (Without Tools): leaderboard
The without-tools track of CAIA (Crypto AI Agent benchmark) — models answer high-stakes crypto and web3 analyst tasks from their own knowledge, with no tool access; Pass@1 (%).
Metric: Pass@1 (%). Source: huggingface.co. Saturation forecast: Around 2030. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 28.1 | #98 |
| 2 | O3 | 22.5 | #124 |
| 3 | DeepSeek R1 | 21.9 | #251 |
| 4 | Grok 4 Fast (Reasoning) | 21.3 | #254 (Grok 4 Fast) |
| 5 | GPT-4.1 | 20.8 | #248 |
| 6 | Gemini 2.5 Pro | 20.2 | #152 |
| 7 | Gemini 2.5 Flash | 20.2 | #245 |
| 8 | Grok 4 | 20.2 | #181 |
| 9 | GPT-4o | 19.1 | #343 |
| 10 | GPT-OSS-120B | 18.5 | #341 |
| 11 | Kimi K2 | 17.4 | #243 |
| 12 | Llama 4 Maverick | 16.9 | #464 |
| 13 | DeepSeek V3.1 | 15.7 | #264 |
| 14 | Claude Opus 4.1 | 15.2 | #140 |
| 15 | Qwen 3 235B A22B | 14.6 | #313 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=caia-pass-1-without-tools · How It Works · Data refreshed daily, snapshot 2026-10-20.