WeChat Tests AI Editing: WeLM Model & Billion-Dollar Compute Bill
APPSO · wechat · 2026-08-23
WeChat is gray-testing an "AI Editing" feature in chat windows, offering 11 presets like beauty filters and text blurring, powered by its custom WeLM model. The analysis highlights WeChat's strategy: using techniques like HiddenDecoding to optimize inference efficiency and curb costs for its 1.4 billion MAU. Tencent's financials reveal a 176% surge in capital expenditure due to AI infrastructure, with new AI products dragging down operating profit by 10.5 billion RMB. WeChat aims to integrate its AI assistant "Xiaowei" across existing scenarios rather than as a standalone app.
More from Infra
- Mistral reportedly plans up to 1 GW of European compute capacity by 2030 — emmanuelvivier · 2026-08-23
- Nvidia hikes AI product prices by over 15% amid rising memory costs — emmanuelvivier · 2026-08-23
- Nvidia hikes some AI product prices by over 15% on surging memory chip costs — emmanuelvivier · 2026-08-23
- Contextual News Search APIs: A Deep Comparison for AI, RAG, and Research — ermanos12 · 2026-08-23
- Qwen3.8-27B MTP Grafted to Unsloth Saves RAM, Requires Thinking Mode — Nyghtbynger · 2026-08-23
- Running Kimi K3 on 8x B300: $190 per million tokens, full cost breakdown — OtherRaisin3426 · 2026-08-23