Particle.news
Download on the App Store

OpenAI and SpaceXAI Open New Frontier Models as U.S. Runs Voluntary Security Checks

Regulators are conducting Commerce Department reviews to probe cybersecurity risks posed by faster, cheaper, full‑duplex voice models.

Overview

  • On Thursday OpenAI expanded public preview access to its GPT‑5.6 family (Sol, Terra, Luna) after cooperating with Commerce Department testing that reviewed the model for cybersecurity risks.
  • OpenAI simultaneously launched GPT‑Live, a full‑duplex voice system that can listen and speak at the same time and delegates complex web search and reasoning tasks to a background frontier model currently identified as GPT‑5.5.
  • SpaceXAI made Grok 4.5 publicly available on Thursday, marketing it as faster and more token‑efficient for coding and agent tasks and with vendor‑sourced claims that it rests on an xAI V9 foundation of roughly 1.5 trillion parameters.
  • Published pricing comparisons show cost and latency have become competitive levers with Grok listed at about $2 per million input tokens and $6 per million output tokens while OpenAI’s lower‑cost Luna is priced around $1 per million input and $6 per million output, changes that will affect enterprise operating costs.
  • The Commerce checks follow recent limits placed on Anthropic and reflect a voluntary executive‑order framework intended to balance national‑security scrutiny with market competition, a process that could influence future access, export rules and how labs stage releases.