VITA-Bench
We show this table for reference; we do not rank on it.
An interactive real-world agent benchmark grounded in practical consumer-service tasks such as delivery, in-store consumption, and online travel workflows.
Benchmark score on VITA-Bench — November 13, 2026
We compile the VITA-Bench rows from benchmark-owner or independent runs and provider self-reports. Qwen3.7 Max leads the table at 47.9%, as reported by Alibaba, followed by Qwen3.7 Plus (45.6%, as reported by Alibaba) and Sage-A1.5 (45.4%, as reported by InternScience). We do not use these results to rank models overall.
Qwen3.7 Max
Alibaba
Reported by AlibabaQwen3.7 Plus
Alibaba
Reported by AlibabaSage-A1.5
InternScience
Reported by InternScience13 modelsAgenticCurrentDisplay onlyUpdated November 13, 2026
Benchmark score table (13 models)
Of 13 results: 9 lab-reported, 4 benchmark-run.
Ranks on this table count the 4 benchmark-run and independent results. A lab-reported result keeps its place by score, without a rank.
- Lab-reported
- 9
- Independent
- 0
- Benchmark-run
- 4
Showing all 13 results, in score order.
| Rank | Model / configuration | Score | Who ran it | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| No rank on this table | Qwen3.7 MaxAlibabaReported by Alibaba | 47.9% | Reported by Alibaba | Not reported | Closed |
| No rank on this table | Qwen3.7 PlusAlibabaReported by Alibaba | 45.6% | Reported by Alibaba | Not reported | Closed |
| No rank on this table | Sage-A1.5InternScienceReported by InternScience | 45.4% | Reported by InternScience | Not reported | Open |
| No rank on this table | Qwen3.6 PlusAlibabaReported by Alibaba | 44.3% | Reported by Alibaba | Not reported | Closed |
| No rank on this table | Qwen3.5 397BAlibabaReported by Alibaba | 43.7% | Reported by Alibaba | Not reported | Open |
| No rank on this table | Agents-A1-4BInternScienceReported by InternScience | 40.3% | Reported by InternScience | Not reported | Open |
| No rank on this table | Agents-A1InternScienceReported by InternScience | 38.8% | Reported by InternScience | Not reported | Open |
| No rank on this table | Qwen3.6-35B-A3BAlibabaReported by Alibaba | 35.6% | Reported by Alibaba | Not reported | Open |
| 1 | Claude Opus 4.5AnthropicRun by VITA-Bench | 23.3% | Run by VITA-Bench | Not reported | Closed |
| No rank on this table | LongCat-Flash-Lite-SparseMeituanReported by Meituan | 21.7% | Reported by Meituan | Not reported | Open |
| 2 | DeepSeek V3.2DeepSeekRun by VITA-Bench | 18.5% | Run by VITA-Bench | Not reported | Open |
| 3 | Claude Sonnet 4.5AnthropicRun by VITA-Bench | 17.0% | Run by VITA-Bench | Not reported | Closed |
| 4 | GLM-4.7Z.AIRun by VITA-Bench | 15.5% | Run by VITA-Bench | Not reported | Open |
A dash in the rank column marks a lab-reported result. It keeps its place by score, without a rank on this table.
- Lab-reported
- Benchmark-run
- Independent
Among the reported VITA-Bench rows, Qwen3.7 Max is first at 47.9%, as reported by Alibaba. The third row is 2.5 points behind. The broader top-10 range is 26.2 points, so the table still separates the published systems.
13 models have been evaluated on VITA-Bench. The benchmark falls in the Agentic category. VITA-Bench is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About VITA-Bench
Year
2025
Tasks
Interactive consumer-service agent tasks
Format
End-to-end interactive agent evaluation
Difficulty
Long-horizon real-world workflows
VITA-Bench is built to test realistic interactive agent behavior rather than toy tool calls. It stresses long-horizon coordination, tool selection, changing user intent, and domain switching across daily-life applications.
Freshness and provenance
Version
VITA-Bench 2025
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does VITA-Bench measure?
An interactive real-world agent benchmark grounded in practical consumer-service tasks such as delivery, in-store consumption, and online travel workflows.
Which model scores highest on VITA-Bench?
Qwen3.7 Max by Alibaba currently leads with a score of 47.9% on VITA-Bench, as reported by Alibaba.
How many models are evaluated on VITA-Bench?
13 AI models have published results on VITA-Bench in the BenchLM catalog.
Compare top models on VITA-Bench
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.