AI research firm DeepSeek has proposed a new mathematical framework to improve the stability and scalability of large language models.
Italy has closed its probe into DeepSeek after reviewing concerns around AI hallucinations, reflecting Europe’s evolving approach to AI oversight.
DeepSeek AI is signalling a shift toward efficiency focused research, aiming to improve AI performance while reducing compute and resource intensity.
DeepSeek launches new reasoning AI models with performance comparable to top global systems, advancing open access innovation.
DeepSeek’s open-source model DeepSeekMath-V2 solves five of six IMO 2025 maths problems and posts top scores on Putnam, pioneering self-verifiable AI reasoning.
China now leads globally in downloads of open-source AI models, marking a shift in the balance of influence in AI development and diffusion.
AWS adds Alibaba’s Qwen3 and DeepSeek V3 to its Bedrock platform, expanding enterprise access to advanced AI models for global business applications.
OpenAI releases open-weight AI models optimized for laptops, marking shift toward open-source competition with Meta, Mistral.
Since its initial launch in January 2025, the original R1 model quickly established itself by offering a performance comparable to renowned Western models at significantly lower costs.
