DeepSeek-V4 Delivers Frontier AI Reasoning at 1/6th Cost of GPT-5.5
DeepSeek released V4, a 1.6-trillion-parameter model delivering frontier AI reasoning at one-sixth the cost of GPT-5.5…
Breaking news, research, and analysis on artificial intelligence — covering the models, methods, and shifts shaping the field.
DeepSeek released V4, a 1.6-trillion-parameter model delivering frontier AI reasoning at one-sixth the cost of GPT-5.5…
DeepSeek released its V4 model offering near state-of-the-art AI coding performance at one-sixth the cost of…
XPENG launched VLA 2.0 as a production-ready autonomous driving system in March 2026, challenging Tesla's beta…
Researchers introduce Train-to-Test scaling laws that optimize AI compute by jointly considering training and inference costs,…
New research challenges assumptions about AI reasoning, showing latent-state dynamics drive capabilities more than chain-of-thought processes.…
UCL researchers have developed a quantum-AI hybrid that delivers unprecedented accuracy in predicting complex systems like…
New AI architecture breakthroughs include Train-to-Test scaling laws that optimize training and inference costs, Google's specialized…
Researchers introduce Train-to-Test scaling laws that optimize AI models by jointly considering training costs and inference…
New research reveals AI reasoning occurs primarily through internal latent-state dynamics rather than visible chain-of-thought processes.…
Claude Opus 4.6 tops new AI benchmark leaderboard with 94.1% on thermodynamic reasoning tests, while Google…
New research reveals that AI models process complex reasoning through internal latent states rather than visible…
Anthropic launched Claude Design, an AI tool that creates professional prototypes from text prompts, setting new…