Sources
AI Explained Dissects the 48-Hour Model Explosion — Sol Ultra, Meta Muse, New Grok — and Argues Price, Not Benchmarks, Is the Story
AI Explained's breakdown of the July 8–9 wave (OpenAI's GPT-5.6 Sol/Terra/Luna, xAI's Grok 4.5, Meta's Muse Spark 1.1) argues the real headline isn't a single benchmark record but collapsing cost: near-frontier capability at a fraction of prior prices. On the Artificial Analysis index the top three cluster within six points (Fable 59.9, Sol 58.9, Grok 54), and Muse Spark 1.1 is a 1M-context agentic model rivaling GPT-5.5/Opus 4.8. The builder takeaway is that model choice is fast becoming a price-per-token decision, not a capability one.
↳ Follow the thread