VITA-Bench

An interactive real-world agent benchmark grounded in practical consumer-service tasks such as delivery, in-store consumption, and online travel workflows.

11 models

Higher is better.

1Qwen3.7 MaxAlibaba47.9
2Qwen3.7 PlusAlibaba45.6
3Qwen3.6 PlusAlibaba44.3
4Qwen3.5 397BAlibaba43.7
5Agents-A1InternScience38.8
6Qwen3.6-35B-A3BAlibaba35.6
7Claude Opus 4.5Anthropic23.3
8LongCat-Flash-Lite-SparseMeituan21.7
9DeepSeek V3.2DeepSeek18.5
10Claude Sonnet 4.5Anthropic17.0
11GLM-4.7Z.AI15.5

Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric: End-to-end interactive agent evaluation
Tasks: Interactive consumer-service agent tasks
Authors: Meituan LongCat Team (2025)
Source: VitaBench: Benchmarking LLM Agents with Versatile Interactive Tasks in Real-world Applications ↗

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.