AI Pulse · Frontier Labs Watch

Bank of America just started running a standing tracker on frontier AI models, built on data from Artificial Analysis. It ranks models on intelligence and cost side by side, with Claude Opus 5 currently first, ahead of Claude Fable 5 and GPT-5.6 Sol. The same dashboard shows token pricing down 9% month over month across the board.

Here’s the part that matters more than the rankings. A major bank building a desk-level tracker for AI model intelligence versus cost is not a research curiosity. It’s a sign that institutional capital now treats model choice as a tradeable input, something you monitor the way you’d monitor commodity prices or currency spreads. When banks start tracking which model gives you the most capability per dollar, they’re treating “which AI to use” as a decision with financial weight, not just a technical one. The original report frames it exactly that way.

For a business owner, the practical takeaway isn’t which model wins this month. That will change. The takeaway is that prices keep dropping while capability keeps climbing, and the gap between models keeps shifting fast enough that a bank felt it needed a dedicated tracker just to keep up. If you’re locked into one model or one vendor because it seemed like the safe pick a year ago, you’re probably paying more and getting less than what’s now available. Treating your AI stack as fixed instead of something you revisit quarterly is starting to look like a real cost.

This is exactly the kind of thing we build into an AI command centre, so you’re not stuck manually tracking which model is worth your money this month. Worth checking your own setup against what’s actually leading right now, not what was leading six months ago.