SpaceXAI and OpenAI shipped competing frontier models within 24 hours of each other. On Wednesday, SpaceXAI, the AI arm of Elon Musk’s newly public SpaceX, released Grok 4.5, its first model since going public, trained specifically for coding and agents. A day later, OpenAI rolled out GPT-5.6, a three-tier family led by its Sol model, after weeks of a government-gated preview. Both companies pitch their models as workhorses for coding, research, and knowledge work, with Anthropic’s Claude Opus invoked as the benchmark. SpaceXAI’s positioning for Grok 4.5 leans on price and speed: at $2 per million input tokens and $6 per million output tokens, it undercuts rivals. Musk described Grok 4.5 as roughly on par with Opus 4.7, which is not Anthropic’s current flagship, as Opus 4.8 and Claude Fable 5 have since superseded it. On DeepSWE 1.1, Grok 4.5 scored 53%, trailing Opus 4.8’s 59% and GPT-5.5’s 67%, with Claude Fable 5 at 70%. On SWE-Bench Pro, Grok 4.5 scored 64.7%, edging out GPT 5.5’s 58.6%, but behind Opus 4.