The New Stack
·
2 days ago
Open-source and open-weight AI models are claimed to be approximately four months behind proprietary frontier models while costing roughly ten times less per token, with Featherless demonstrating annual costs of $90,000 for their optimized GLM 5.2 model versus $1.5 million for GPT-5.5 or Claude Opus at 100 billion monthly tokens. The cost comparison shows that enterprises paying for proprietary models are primarily paying for enterprise wrapper features like integration and observability rather than raw model intelligence. As open-source models improve and approach frontier capability levels, enterprises may find themselves locked into expensive proprietary contracts while open alternatives become increasingly viable alternatives.
Ben's Bites
·
2 days ago
OpenAI released GPT-5.6 with three models (Luna, Terra, Sol) featuring five thinking levels plus a new Ultra mode, and merged the ChatGPT and Codex apps into a single application with a new ChatGPT Work mode. The three models have different strengths with Luna suited for simple tasks, Terra offering minor improvements over GPT-5.5, and Sol excelling at UI and writing, while usage limits are consumed significantly faster at higher thinking levels. Users should adjust their model and thinking level selections based on task complexity to avoid rapidly depleting their weekly usage allowances.
Sifted
·
2 days ago
NestAI, founded by Peter Sarlin, released its first AI models designed specifically for military use, aiming to reduce Europe's dependence on foreign defence technology providers. The company raised €100 million in November, has grown to 200 people, and is running pilot programs with Estonian and Finnish armed forces that enable autonomous drone operations and battlefield mission orchestration. The launch addresses European security concerns highlighted by recent US export restrictions on Anthropic's models, with NestAI focusing on domain-specific military applications rather than competing with general-purpose frontier AI models.
