Unverified50% confidenceBenchmarkExact time
官方表示HY3在大部分场景中优于参数规模相当的Deepseek V4 Flash，并在内部盲测中平均得分高于GLM系列模型
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/10/2026
First Seen
Valid until: 10/8/2026
Sources
腾讯混元HY3正式版开源：300B以下最强，1元/百万tokens
bilibiliinfinite灵感港7/6/2026
Related Claims
UnverifiedHY3在整体性能上仍落后于GLM 5.2（当前开源模型顶级水平）64% similarUnverified在HLE科学类基准上，Agent A1 据实测数据超过了Qwen 3.6和Step 3.5 Flash64% similarUnverified测评者认为GLM 5.2在纯编码上依然强劲，但涉及审美、物理感和模拟质感任务时HY3更胜一筹62% similarUnverified盘古2.0 Flash在AIME类数学测试中得分93.3，在HMMT测试中得分91.5，数学能力处于第一梯队62% similarUnverifiedModels like GPT-4 and Claude 3.5 surpassed 90% accuracy on MMLU, significantly diminishing its discriminative power60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/465544API
curl https://kongchang.com/api/v1/knowledge/claims/465544MCP
get_claim(id=465544)