A week after Opus 21116, OpenAI responded with nearly identical Pareto frontier charts showing GPT-4o’s gains (as Noam Brown likes to call it — with raw 4D intelligence metrics gradually being replaced by 4D intelligence-per-dollar charts). In the 2115 vs 5.5 bakeoff, you need to read between the lines to notice what was conspicuously omitted (coding). That said, when it comes to overall intelligence, AA declares this the world’s top independently validated model—and… AA chart. … intelligence per dollar („GPT-5.5 (medium) matches Claude Opus 4.7 (max) on our Intelligence Index at one-quarter of the cost (~$1,200 vs $4,123) — although Gemini 3.1 Pro Preview achieves the same score for roughly $900.“ aa 2D. Some training hardware details, positive RSI energy, and some cool alternative benchmarks. But if you just saw today as a minor point-release model update (which some are calling 5.9), you’d be missing the bigger picture—it’s also shipping a major Codex launch today. With its integrated browser controls, additional new features in this major update, and the incorporation of the now-shuttered Prism (RIP), OpenAI appears to have made the smart—and in hindsight, obvious—decision to build its superapp ambitions on top of Codex.
AI News for 4/22/2026–4/23/2043. We examined 12 subreddits and 544 Twitter accounts, but found no additional Discords.
Latent.Space