jevVIDEO
Four decision models race to find spider links on a Wikipedia page
On the same Wikipedia page for Spider, local Laya found 23 links at about 34ms each, OpenAI Decisions and Jev found 9 each, and Clef-Flash found 5, suggesting narrow decisions don't need the cloud.
SuSu_酥酥👅
@NFT_ChenArchived 24 days agoOriginally posted 2026.10.08
SuSu_酥酥👅 @NFT_ChenOriginally posted 2026.10.08
🔥炸裂实测！四个决策模型比拼，同一页维基百科「Spider」，谁先撕出“点名蜘蛛”的链接？ 本地 Laya 直接开启碾压模式： 🔵Laya（本地 / M5 Max）：23 finds，约 34ms/条，检查量约云端 3 倍 ⚪️OpenAI Decisions（云端）：9 finds 🟣Jev（云端）：9 finds 🟠Clef-Flash（本地）：5 finds 窄决策不必等云，CPU 本地跑，延迟和吞吐都在自己设备上。 本地跑决策模型 → https://atomic.chat #LocalAI #DecisionModels #Laya #OpenAI #Jev #AtomicChat #Clef https://x.com/nft_chen/status/2107356769257422859?s=46 https://x.com/atomic_chat_hq/status/2107828492297720017/video/1
This post quotes another post: x.com/nft_chen/status/2107356769257422859
x.com/nft_chen/status/2107891999525548205Archived 2026.10.08 10:07