Qwen 3.5 Plus: benchmark results
Alibaba's flagship Qwen 3.5 Plus API tier with thinking and non-thinking modes. Provider: Alibaba. Released 2026-02-16. Access: API.
Aggregate Rating 1741 ± 4, rank #125 of 1656 rated models, from 783 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AirGroundBench - Bearing Perception | 48.28 | Accuracy (%) on the bearing-perception VQA task; higher is b | 100 |
| AirGroundBench - Counterfactual Reasoning | 48.57 | Accuracy (%) on the counterfactual spatial-reasoning VQA tas | 100 |
| AirGroundBench - Scale Alignment | 59.57 | Accuracy (%) on the cross-view scale-alignment VQA task; hig | 100 |
| CLBench-V - Information Application (L1) | 29.54 | Score (%; instance-weighted mean of the dataset scores, each | 100 |
| Cloud-OpsBench (TrainTicket) - Evidence Closure Rate | 25 | Evidence closure rate (%): share of episodes whose trajector | 100 |
| Cloud-OpsBench (TrainTicket) - Evidence-Order Consistency | 58 | Evidence-order consistency (%): mean share of milestone grou | 100 |
| Cloud-OpsBench (TrainTicket) - Fault-Type Accuracy | 79 | Fault-type accuracy (%): share of episodes whose predicted f | 100 |
| Cloud-OpsBench (TrainTicket) - Milestone Coverage | 64 | Milestone coverage (%): mean share of the case's expert diag | 100 |
| CodeRouterBench - LLM Router (0-Shot) | 46.87 | Average task score (%) of the backend the model routes each | 100 |
| EgoCoT-Bench - Local Cause-Outcome Inference | 74.72 | Answer accuracy (%, four-way multiple choice, strict exact m | 100 |
| EgoCoT-Bench - Predictive and Causal Inference | 78.21 | Answer accuracy (%, four-way multiple choice, strict exact m | 100 |
| EgoCoT-Bench - Progress Reasoning | 81.12 | Answer accuracy (%, four-way multiple choice, strict exact m | 100 |
Interactive version: theaggregate.ai/model?slug=qwen-3-5-plus · How It Works · Data refreshed daily, snapshot 2026-10-20.