OpenAI and Broadcom have introduced Jalapeño, a custom chip for LLM inference designed to improve performance and efficiency in production deployment of language models.
Frontier LLMs solve fewer than one-third of 87 multi-GPU CUDA benchmark tasks, though some generated kernels still outperform public reference implementations.
OpenAI offers specialized AI tools for automated detection, validation, and remediation of security vulnerabilities to scale vulnerability management in large organizations.
ChatGPT retains the most users but must cede market share to Gemini (27.7%) and Claude (10.3%), as the industry shifts from user acquisition to monetization.
Fully utilized flat-rate subscriptions for ChatGPT and Claude generate costs that exceed monthly fees by a multiple, forcing providers to rely on low average utilization rates while enterprise customers increasingly switch to cheaper alternatives or in-house systems.