Alibaba's new Qwen 3.8 27B vision AI model runs on consumer laptops as a 17GB file, delivering impressive code generation and image analysis capabilities, but its default 'xhigh' reasoning setting causes excessive overthinking on simple tasks, and speed remains limited at 15–30 tokens per second — though a 72% boost …
A new open-source tool called 'Soup' is revolutionizing AI development by allowing developers to fine-tune massive 8-billion parameter language models on a standard 4GB laptop GPU using a single command, thanks to a breakthrough 'layer streaming' technique that achieves 119.6 tokens per second at just 3.32 GB peak VRAM.
Google launches Gemini 3.7 Flash with a 50% price cut just three weeks after its predecessor, boasting major coding and workflow improvements, but internal leadership turmoil, talent departures, and delays to its flagship Pro model raise serious questions about whether Google can reclaim dominance in an increasingly competitive AI race.
Google launches Gemini 3.7 Flash with significant benchmark gains in coding and specialized fields, but at $0.75 per million input tokens, it remains pricier than OpenAI's competing GPT-5.6 Luna — all while navigating major leadership shakeups at DeepMind.
Google DeepMind launches Gemini 3.7 Flash, a powerful new AI model boasting enhanced reasoning, 1M token context support, and improved performance in coding and enterprise workflows, all at a competitive price of $0.75 per 1M input tokens through December 2026.
MiniMax Music 3.0 launches as a powerful open-weights AI model that generates complete, fully arranged songs up to five minutes long from a single text prompt, giving creators unprecedented control over instrumentation, vocals, and song structure with no music production expertise required.
DeepSeek launches DeepSeek Harness, an open-source AI agent framework built on a plugin-based architecture, already exploding in popularity with over 84,000 GitHub stars and 7,400 forks since its release.
WPP slashes humanoid robot training time from 24 hours to under one hour using NVIDIA RTX PRO 6000 Blackwell GPUs on Google Cloud, while a cutting-edge motion capture and digital twin workflow enables robots to mimic natural human movement with remarkable stability.
An unreleased Anthropic AI model makes historic progress on the 150-year-old Riemann Hypothesis, autonomously running for 36 hours, testing 650 ideas across 60 subagents, and successfully pushing the lower bound of verified solutions — a breakthrough confirmed by mathematicians and formalized in Lean, sparking fierce debate about AI's role in …
DeepSeek silently launches its flagship V4 Pro model, updated to DeepSeek-V4-Pro-0813, nearly matching Claude Fable 5 on agent benchmarks at just $0.65 per million tokens — 46 times cheaper than Claude's $30 rate — with MIT-licensed weights publicly available on Hugging Face for independent verification.