DeepSeek V4 Pro
DeepSeek releases V4 Pro, a 1.6-trillion-parameter Mixture-of-Experts model with 49 billion active parameters per token, achieving a top open-weight score of 80.6% on SWE-bench Verified. The model requires multi-H200/H10…
AI Products news and analysis on Web Pulse: 25895 curated articles tracking the latest AI Products developments, tools, and research, updated continuously from vetted sources.
DeepSeek releases V4 Pro, a 1.6-trillion-parameter Mixture-of-Experts model with 49 billion active parameters per token, achieving a top open-weight score of 80.6% on SWE-bench Verified. The model requires multi-H200/H10…
Anthropic's Claude Fable 5 was taken offline by the US government on June 12 after a reported jailbreak, then returned on July 1 with a stricter safety classifier. Benchmark tests by KiloBench show that before the ban, 2…
Apple's iOS 27 introduces AI-powered features to the Mail app, including Siri integration for summarizing emails and drafting messages, Smart Replies tailored to user writing style, and contextual suggestions for flights…
Researchers have found that large language models (LLMs) can generate diverse software code, potentially reducing common-mode failures, but reliability improvements depend on programming language and generation settings.…
Meta launched its first AI image generator, Muse Image, which automatically opts in public Instagram accounts, allowing anyone to use users' photos in AI-generated images unless they manually opt out. The feature raises …
Researchers introduced Subspace-Aligned Rewiring (SAR), a post-training editing method for large language models that isolates reasoning-effective components in spectral space, preserving over 99% of performance using as…
Researchers have developed CONFLUX, a 3D medical imaging model combining a latent diffusion model with reinforcement learning to generate chest CT scans with unprecedented precision. The model outperforms existing benchm…
GitHub Copilot added Kimi K2.7 Code, Moonshot AI's open-weight coding model, on July 1 for Pro, Pro+, and Max subscribers, expanding to Business and Enterprise on July 7. It is the first open-weight model in Copilot, all…
An engineer running multi-region AI workloads reports that Chinese LLM APIs from DeepSeek, Qwen, GLM, and Kimi offer dramatically lower costs than US providers like OpenAI, Anthropic, and Google, with DeepSeek V4 Flash c…
Together AI and AppsFlyer top Green Flag Digital's Q3 2026 ranking of private tech companies, which weighs funding, web traffic, and brand search volume. Together AI leads on the strength of an $800M funding round and co…
An online insurance agency owner built an MCP server that lets AI agents request disability insurance quotes, using Claude Code despite not being a developer. The server offers six read-only tools and one action that wri…
Insilico Medicine's AI-discovered lung fibrosis drug, Rentosertib, has entered Phase III clinical trials for idiopathic pulmonary fibrosis, marking a rare late-stage test for AI drug discovery. The company claims it is t…
Founders must now optimize for AI chatbots like ChatGPT rather than traditional search engines, as Gartner predicts a 25% drop in search volume by 2026 and Google's AI Overviews produce zero-click results for 80-83% of q…
Apple sued OpenAI and two former employees on Friday, alleging theft of trade secrets related to consumer hardware. The complaint claims former Apple engineer Chang Liu accessed confidential files after leaving the compa…
China is advancing its global soft power by releasing open-weight AI models that are widely adopted in Africa and other regions, while the U.S. relies on closed-source proprietary models. Andrew Ng warns that this trend,…
Apple filed a lawsuit against OpenAI and two former employees, including ex-Apple designer Tang Tan, accusing them of stealing trade secrets. Apple alleges Tan misused confidential information to benefit OpenAI and recru…
Cognition Labs launched Devin, the 'first AI software engineer,' in March 2024 with a benchmark score of 13.86% on SWE-bench, but independent evaluations later showed it failed 14 out of 20 real tasks. Despite a partly s…
Samsung users report that Bixby's voice randomly changes after a recent update, switching between the selected voice and an unfamiliar lower voice. The Bixby team acknowledged the complaint but did not confirm a bug or p…
A developer compares Mem0 and TurboMem memory layers for TypeScript AI agents, highlighting architectural trade-offs between service-based and embedded memory. Slack Engineering introduces 'Agentic Testing,' an AI-driven…
A developer built Sonic Kinetic, a workout timer that uses Google Gemini to generate interval routines and ElevenLabs to produce real-time audio coaching with variable intensity. The Go backend sends workout parameters t…