Unverified85% confidenceFactTime unknown
Qwen3.5在IFBench指令遵循评测中得分76.5，超过GPT-5.2的75.4和Claude 4.5 Opus的58
1
Sources
85%
Confidence
Medium-term (~90 days)
Relevance
6/1/2026
First Seen
Valid until: 8/30/2026
Sources
Qwen3.5深度解析：混合注意力架构实现19倍长上下文加速
bilibili雪天鱼
Related Entities
Related Claims
UnverifiedQwen3.6 35B在SWE Bench Verified上的成绩距离Claude Opus 4.6仅差1.6个百分点77% similarUnverifiedQwen3.5在AIME 2026数学竞赛评测中得分91.3，GPT-5.2为96.777% similarUnverifiedQwen3-4B在多项推理类评测指标上已与DeepSeek V3或GPT-4o相差无几77% similarUnverifiedQwen3.5在HLE博士级综合题评测中得分28.7，而GPT-5.2为35.576% similarUnverifiedQwen 3's flagship version matches or surpasses GPT-4o and Claude 3.5 Sonnet in multiple benchmark tests.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/18080API
curl https://kongchang.com/api/v1/knowledge/claims/18080MCP
get_claim(id=18080)