Unverified50% confidenceBenchmarkExact time
Mixar在LOD生成测试中生成LOD0/1/2三个层级，贴图从4K降至2K和1K，共创建52张新贴图，结果符合预期
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/8/2026
First Seen
Valid until: 10/6/2026
Sources
Mixar实测：AI原生Blender分支版，烘焙与LOD一句话搞定
bilibiliodOObo7/7/2026
Related Claims
Unverified在27B混合架构、4B稠密模型、36B MoE三类模型上以8k上下文测试，Prefill阶段性能无变化，Decode阶段打补丁后快约1.4%66% similarUnverified基准测试中每个仓库每一边各跑4次取中位数62% similarUnverified混合精度训练中，FP16 每个参数占 2 字节显存，FP32 每个参数占 4 字节显存61% similarUnverifiedAVX-512指令可在单条指令内完成16个float32的乘加运算，相较标量循环带来十余倍吞吐提升60% similarUnverified叠加cv=5的5折交叉验证后，144种参数组合需训练144×5=720次模型60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/261200API
curl https://kongchang.com/api/v1/knowledge/claims/261200MCP
get_claim(id=261200)