Four American Labs Now Dominate the AI Frontier

In a single week, the AI landscape shifted from a two-company race to a four-way competition. On July 8th, Elon Musk announced Grok 4.5. The next day, OpenAI released GPT-5.6 across three tiers — Sol, Terra, and Luna — with Luna reportedly post-trained by the higher-end Sol model in what OpenAI called an experiment in recursive self-improvement. Meta then showed signs of life with Muse Spark. Combined with Anthropic's Fable 5, which had been released just days earlier on July 2nd, the result is what Alex Wiesner-Gross described as "a convergence episode": intelligence getting cheaper, interfaces becoming more human, and AI acquiring tools and hands — all in one week.

The Frontier Opens Up

The most important structural change is that the frontier is no longer a duopoly. Alex put it plainly: "We went from a beginning of the week where the frontier was a duopoly between Anthropic and OpenAI — if you want to be on the frontier, you have to be using models from one of those two vendors — and now we have four American labs at the optimal frontier." Those four are Anthropic, OpenAI, Meta, and SpaceX AI. Google, notably, is not on the optimal frontier; Gemini 3.5 Pro is reportedly delayed.