Qwen 3.5 Plus (Thinking): benchmark results

Qwen 3.5 Plus evaluated with thinking enabled. Provider: Alibaba. Released 2026-02-16. Access: API.

Unified ELO 1649 ± 1, rank #248 of 1761 rated models, from 26 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
SecCodeBench67.08Total Score97.1
BenchTable72.7Total Score (%)89.5
Vals AI MedQA95.21Accuracy (%)87.2
Vals AI LegalBench85.1Accuracy (%)84.4
Vals AI CorpFin v265.31Accuracy (%)77.4
Vals AI LiveCodeBench85.33Accuracy (%)76.8
Vals AI MMLU-Pro87.18Accuracy (%)75.9
LLM2014 Logic 2026-0253.8Median Score75.6
Vals AI GPQA87.37Accuracy (%)72.6
LLM2014 Logic 2026-0351.81Median Score70.7
CLBench19.8Solving Rate (%)68.6
Vals AI AIME86.04Accuracy (%)65.8

Interactive version: theaggregate.ai/model?slug=qwen-3-5-plus-thinking · How It Works · Data refreshed daily, snapshot 2026-09-05.