VitaBench: leaderboard
66-tool agent benchmark across food delivery, retail, and travel domains with 100 cross-scenario and 300 single-scenario tasks. Even the best model reaches only 62% on single-scenario and 32.5% on cross-scenario.
Metric: Cross-Scenario Avg@4 (%). Source: vitabench.github.io. Saturation forecast: Around August 2028. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (High) | 32.5 | #99 (Gemini 3 Flash) |
| 2 | Gemini 3 Pro (High) | 31.5 | #80 (Gemini 3 Pro) |
| 3 | Gemini 3 Pro (Low) | 30 | #80 (Gemini 3 Pro) |
| 4 | LongCat-Flash-Thinking-2601 | 29.3 | #229 |
| 5 | Claude Opus 4.5 | 28.5 | #83 |
| 6 | O3 (High) | 26.3 | #124 (O3) |
| 7 | GPT-5.2 (xHigh) | 24.3 | #110 (GPT-5.2) |
| 8 | DeepSeek V3.2 | 24 | #205 |
| 9 | Claude Sonnet 4.5 | 23.5 | #142 |
| 10 | LongCat-Flash-Chat | 22.8 | #331 |
| 11 | O4 Mini (High) | 19.5 | #180 (O4 Mini) |
| 12 | GLM-4.7 | 18.3 | #192 |
| 13 | Qwen 3 235B A22B 2507 (Thinking) | 14.5 | #260 (Qwen 3 235B A22B 2507) |
| 14 | Qwen 3 Max | 14.3 | #208 |
| 15 | Seed 1.8 | 13.8 | #139 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=vitabench · How It Works · Data refreshed daily, snapshot 2026-10-20.