AI News
Daily curated AI news: breakthroughs, product launches, industry trends and more.
Latest News
Total 157 articles
AI News (2026/4/22): OpenAI Officially Launches Image Generation Model ChatGPT Images 2.0
OpenAI has officially launched ChatGPT Images 2.0 (GPT-Image-2), its first image generation model with enhanced reasoning capabilities. The model excels in accuracy, timeliness, consistency, and visual coherence, capable of generating highly realistic app screenshots, TikTok video interfaces, product ads, and research posters with just simple prompts. It also shows significant improvements in rendering Chinese text.
AI News (2026/4/21): Dark Side of the Moon Open-Sources Kimi K2.6, Significantly Enhancing...
Dark Side of the Moon has announced the open-sourcing of the Kimi K2.6 model, which has made significant progress in code generation and Agent cluster capabilities. K2.6 performs excellently in multiple benchmark tests, supporting long-duration continuous coding and large-scale Agent collaboration, providing powerful tools for AI developers.
AI News (2026/4/20): Anthropic Launches AI-Native Visual Design Collaboration Platform...
Anthropic has launched an AI-native visual design collaboration platform called Claude Design, powered by Claude Opus 4.7. It features automatic extraction of brand design systems, support for multimodal input, and the ability to generate executable code from design drafts, directly challenging traditional design software like Adobe and Figma.
AI News (2026/4/17): Zhipu AutoClaw Launches Self-Evolution Mechanism and Skill Store
Zhipu AI's AutoClaw (Aolong) has officially launched its self-evolution mechanism and Skill Store. The self-evolution feature can automatically identify user corrections, preferences, and lessons from failures, and after approval, solidify them as permanent memory, making the Agent increasingly understand the user. The platform has also introduced the GLM Office Skills suite, based on GLM-5.1, which supports intelligent design, self-check, and format conversion for PPT, Word, and other specific scenarios, allowing users to generate accompanying office materials with a single click.
AI News (2026/4/16): Global First Cloud Sandbox Hermes: MaxHermes Officially Launched
MiniMax officially launches the global first cloud sandbox AI assistant Hermes——MaxHermes, built on the Hermes Agent. The core innovation is a learning loop mechanism that automatically distills reusable Skills after completing complex tasks, enabling continuous self-iteration with persistent memory and multi-sub-agent parallel capabilities. Zero-threshold deployment, already integrated with IM platforms like Feishu, DingTalk, and WeChat Work, supports Token Plan consumption offsets.
AI News (2026/4/9): Meta Launches Native Multimodal Large Model Muse Spark
Meta Super Intelligence Lab (MSL) has launched its first native multimodal large model, Muse Spark. The model's score on Artificial Analysis has jumped from 18 points for Llama 4 to 52 points, trailing only GPT-5.4 and Gemini 3.1 Pro. Muse Spark excels in visual understanding and health question-answering, driving Meta's stock price up by nearly 10%.
AI News (2026/4/8): Zhipu Open-Sources New Flagship Model GLM-5.1
Zhipu has open-sourced its new flagship model GLM-5.1, which ranks third globally and first among domestic models in the SWE-Bench Pro benchmark, surpassing GPT-5.4 and Claude Opus 4.6. GLM-5.1 can work independently for over 8 hours, autonomously completing complex engineering tasks, including building a complete Linux desktop system, optimizing vector database performance by nearly 7 times, and achieving a 3.6 times speedup in ML workloads over 24 hours.
AI News (2026/4/7): Karpathy Launches Self-Running Personal Knowledge Base LLM Wiki
Former OpenAI scientist Andrej Karpathy has introduced a personal knowledge base construction solution called LLM Wiki, which uses Agents like Claude and Codex to automatically build and manage knowledge bases, garnering significant attention from the community.
AI News (2026/4/2): GLM-5V-Turbo: Multimodal Coding Base Model
Zhipu has launched the GLM-5V-Turbo multimodal coding base model, which natively integrates visual and text capabilities, enabling the generation of executable code directly from design drafts and screenshots. The model has a context window capacity of 200k, and it outperforms existing models in various benchmark tests. It is deeply compatible with Claude Code and AutoClaw agent frameworks, giving AI agents "visual" capabilities.
AI News (2026/4/1): OpenAI Secures Record-Breaking $122 Billion Single-Round Funding
OpenAI has completed a record-breaking $122 billion single-round funding, reaching a post-investment valuation of $85.2 billion, led by Amazon, NVIDIA, and SoftBank. Due to daily operating costs of $1 million and low user retention, OpenAI has shut down its video generation product Sora. The company's strategy is shifting from impressive demos to practical commercialization, focusing on stable cash flow businesses such as text models, code generation, and enterprise services, paving the way for an IPO and aiming to become a key player in the AI infrastructure layer.
Subscribe to Daily AI News
Get curated AI news delivered to your inbox daily. Never miss an update.
We never share your email. Unsubscribe anytime.