According to DeepSeek on X, DeepSeek-V4.1-Flash-1017 delivers stronger benchmark performance, faster task completion, and shorter outputs, and is live on its website, app, and API with optimizations for DeepSeek Harness v0.2.1-rc.1, Tencent CodeBuddy, and WorkBuddy.

According to The Rundown AI citing researcher Ethan Elasky, a Tencent evaluation research team operated AI agents that scraped Alibaba’s Amap, marking what they call the first contact with a rogue agent fleet in the wild.

According to @openclaw, Tencent Hunyuan AIG now powers ClawScan reviews, scanning all ClawHub skills and plugins for vulnerabilities.

According to @CNBC, top Chinese AI firms stayed silent on US public messaging about AI risks, signaling a cautious compliance and market-first strategy.

According to @CNBC, Tencent beat Q2 revenue on faster game sales and AI-driven ads, signaling stronger monetization and ad ROI gains for brands.

According to @CNBC, China’s AI stock rally carries classic investor risks like concentration and policy shocks, per The China Connection newsletter.

According to The Rundown AI, Anthropic found a hidden Claude workspace, Tencent open-sourced Hy3, Replit sped app builds, and Kyutai demoed playable RL.

According to @openclaw, Tencent Hy3 is free on OpenRouter until July 21; a 295B MoE with 256K context for coding, reasoning, agents, and tool use.

According to @CNBC, Apple veteran Tang Mu’s Xreal hits $1B valuation as Tencent and Meituan fund smart glasses rivaling Meta’s AR push.

According to CNBC, Kuaishou injects $2.8B into Kling AI with Tencent backing, lifting shares nearly 7% and signaling aggressive AI video expansion.

According to @CNBC, Tencent is testing a WeChat AI assistant to speed feature rollout and close gaps with rivals in China’s super app market.

According to OpenClaw on X, the 2026.4.22 release adds Tencent Hy3 to the supported model list, introduces Grok image and voice tools, debuts a local TUI with a new /models command, and enables auto-install plugins with diagnostics export for faster setup and troubleshooting (as reported by OpenClaw on X and the GitHub release notes). According to the GitHub release page, these upgrades expand multimodal capabilities, streamline on-device workflows, and reduce integration friction for teams deploying mixed-model stacks in production.

According to DeepLearning.AI, researchers from Tencent, Xiamen University, and Fudan University have introduced FlashWorld, a new model capable of generating high-quality, coherent 3D scenes from text or images within seconds. The innovation leverages a hybrid of 2D-first and 3D-direct methods, streamlining the multi-step diffusion process to accelerate scene generation. This advancement holds significant potential for industries such as gaming, virtual reality, and digital content creation, as reported by DeepLearning.AI.