Unverified50% confidenceBenchmarkExact time
盘古2.0 Flash在GPQA Diamond科学推理测试中（Thinking模式）得分83.7，低于千问3.6 35B的86、27B稠密模型的87.8和DeepSeek V4 Flash的88.1
1
Sources
50%
Confidence
Long-term
Relevance
7/10/2026
First Seen
Sources
华为盘古2.0-Flash横评：指令遵循第一，代码工程垫底
bilibiliAI观模者6/30/2026
Related Claims
Unverified盘古2.0 Flash采用MoE（混合专家）架构，参数规模为92B A6B（总参数92B，激活参数6B）75% similarUnverifiedStep 3.7 Flash 在 V* 基准测试中达到 95.3 分74% similarVerifiedFlash Attention 2在A100 GPU上可达到理论峰值FLOPS的72%73% similarUnverifiedDeepSeek 的模型 V4 Flash 0731 在 Terminal-Bench 2.1 上取得了 82.7% 的成绩69% similarUnverified在DeepSeek V4 Flash部署中DiSpark使单用户生成速度提升60%到85%，Pro系列提升57%到78%，并保持同等推理质量67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/463521API
curl https://kongchang.com/api/v1/knowledge/claims/463521MCP
get_claim(id=463521)