Latest news from the AI world
August 25, 2026: the Jalapeño chip delivers its first measured numbers, the WebMCP Challenge brings together six platforms, Perplexity runs its agent locally.
August 24, 2026: Wan 3.0 produces 30 seconds in one pass, NVIDIA measures peaks at 30x work per watt, Mistral signs with HUMAIN for sovereign AI.
23 August 2026: the creator of Claude Code places Claude on three levels, Gemini 3.7 Flash enters Google Search, Gemma passes one billion downloads.
August 22, 2026: Claude Code Remote Control starts sessions from the phone, GLM-5.3 beats GPT-5.6 Sol and Claude Fable 5 on DeepSWE, NotebookLM available to everyone, Perplexity details its Brain memory.
August 21, 2026: computer use and Skills API go GA on Claude Platform, GitHub Copilot arrives in Slack and Microsoft Teams, DeepSeek unveils its first multimodal model, and Pika Labs launches an ultra-fast TTS.
August 20, 2026: GitHub details the 7h47 outage of August 17 and speeds up its reliability work, Meta AI launches WildArtifactBench, Pika unveils its audio lineup.
August 19, 2026: Claude autonomously designs proteins with Adaptyv Bio, Gemini launches a global student offer, and Replit launches Free Mode with GPT-5.6 Luna.
August 18, 2026: OpenAI slows its models amid Astra cyber risks, Warp launches Warp Factories, and Claude Code launches the Design skill in preview.
August 17, 2026: OpenAI and NVIDIA secure an 8 GW-IT AI site in Ohio, Cursor launches its Origin platform, DeepSeek V4 Pro rolls out at Together AI and Novita, and GitHub details its agentic canvases.
August 14 to 16, 2026: Cursor officially joins SpaceX, Z.ai launches GLM-5.3 with a cyber defense focus, Pika Labs enters the generated audio market with four models, and Claude Code autonomously maintains internal applications at Anthropic.
August 13, 2026: DeepSeek launches V4-Pro in open weights under the MIT license, Google introduces Gemini 3.7 Flash for code and agents (also available in GitHub Copilot), and OpenAI previews Ultrafast, a mode that can reach 14 times standard speed for GPT-5.6 Sol thanks to Cerebras.
August 12, 2026: NVIDIA announces financing partnerships worth more than $500 billion for its AI infrastructure, Alibaba releases the open weights for Qwen3.8-2.4T-A95B, Zed unveils Delta for coding with agents, Google DeepMind launches SL2T on Pixel 11, and Suno signs a global partnership with BMG.