Unverified50% confidenceBenchmarkExact time
在16个单轮任务上共进行314次对比，对阵Gemini Flash（274次）为62胜146平66负，对阵Gemini Pro（40次）为35胜0平5负
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/29/2026
First Seen
Valid until: 11/27/2026
Sources
Related Entities
Related Claims
UnverifiedGemini 3.5 Flash在Humanities Last Exam中得分40.2%，落后于Opus的46.9%和Pro的44.4%69% similarUnverifiedGemini 3 Pro在五轮测试中总共消耗了22.3万个Token，是四个模型中最多的69% similarUnverified在LM Arena的体素（Voxel）生成测试中，7次随机匹配有6次抽到了新版本的Gemini Flash69% similarUnverifiedGemini 3.5 Flash在Agent任务基准测试中获得83.6%的得分69% similarUnverifiedGemini 3.5 Flash支持1,048,576个输入token和65,536个最大输出token67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/820818API
curl https://kongchang.com/api/v1/knowledge/claims/820818MCP
get_claim(id=820818)