OpenAI has published the first performance results for its custom inference ASIC, codenamed 'Jalapeño.' The benchmarks indicate the chip delivers 1.5x to 1.9x more work per watt and 1.7x to 3.6x lower latency than Nvidia's GB200 and GB300 architectures. While volume production isn't expected until 2027, the results signal a major shift as frontier AI labs move to own the full hardware-software stack for inference at scale.
Google and Anthropic are shifting focus from general-purpose chatbots to industry-specific 'centrality.' Google's launch of Gemini Enterprise for Legal and Financial Services, combined with Anthropic's 'Multiplayer AI' strategy, highlights a trend where AI agents are integrated directly into professional workflows and organizational memory rather than acting as standalone assistants.
Anthropic's internal 'Model 2' is reportedly more capable than its flagship Claude Mythos 5, but the company has withheld its release due to uncompleted safety evaluations.
Which OpenAI model family was officially retired from ChatGPT today, ending a 90-day sunset period?