Media reports

Matching articlesArticles60
Browse by monthAll months

Integrating SGLang into Spyre in Under a Week with AI

A two-person team at IBM used AI coding agents to integrate SGLang into the Spyre accelerator in under a week with a ~1,500-line plugin, achieving 3.6× the throughput of reference PyTorch inference and performance on par with vLLM.

Media reportsReportingPyTorch

[AINews] D-Day (Decision Models Day)

Google launched Gemini Decision 1 and the Decisions API for single-pass closed-set decisions, supporting up to 1,000 questions per request, 256K context, 150–250 ms end-to-end latency, and $0.10 per 1M input tokens.

Media reportsReportingLatent Space