Mongolian startup Egune AI pioneers culturally-aware language models for Mongolian language and nomadic traditions, challenging reliance on American and Chinese tech giants despite limited resources.
Google announces its AI prompts now consume minimal energy, with the median prompt using just 0.24 watt-hours of electricity, producing 0.03 grams of carbon dioxide emissions, and consuming 0.26 milliliters of water, a 33-fold reduction in energy use compared to a year ago.
DeepSeek unveils V3.1, a GPT-5 rival optimized for Chinese chips, matching OpenAI's model on benchmarks while undercutting pricing, fueling the AI race between the U.S. and China.
Google unveils AI Inference Impact Metrics, a methodology quantifying the energy, emissions, and water footprint of AI inference tasks like Gemini text prompts, revealing median energy use of 0.24 watt-hours and emissions of 0.03 grams of carbon dioxide per prompt.
Cohere's Command A Reasoning, a highly secure and efficient AI model, outshines privately deployable alternatives, powering the company's flagship enterprise platform North for secure deployment of AI agents and automations.
OpenAI unveils GPT-6, boasting enhanced memory for personalized responses, allowing users to customize the AI's tone and personality for an ideologically neutral or tailored experience.
NVIDIA unveils Granary, a groundbreaking dataset comprising 1 million hours of audio, propelling multilingual speech AI; Apple's Siri gains cross-app voice command capabilities; a guide empowers beginners to master voice cloning in just 11 minutes.
Alibaba's Qwen-Image Edit, an open-source AI model, enables seamless text-to-image edits rivaling Photoshop, from object transformations and style transfers to precise appearance tweaks, offering enterprises cost-effective image editing capabilities.
DeepSeek unveils DeepSeek V3.1, a 685-billion parameter open-source AI model rivaling OpenAI and Anthropic, integrating chat, reasoning, and coding into one model with global access and breakthrough performance at a fraction of the cost.
Anthropic unleashes Claude Sonnet 4 with 1 million token context window at higher pricing; Bessemer identifies two AI startup classes: fast 'Supernovas' and sustainable 'Shooting Stars'; security risks emerge in Model Context Protocol adoption, necessitating new threat models.