The Artificial Analysis Intelligence Index now has a companion tool that does the one thing most AI procurement teams struggle with: plotting raw capability against blended API price on a single chart. Models that sit on the value frontier — where nothing cheaper is also smarter — are the only rational choices. Everything else is dominated, meaning a cheaper model matches or beats it on both axes. The tool's design is deceptively simple. It pulls daily data via the Artificial Analysis free API using a GitHub Actions cron job, blends input and output token pricing at a 3:1 ratio, and plots each model's Intelligence Index score against that blended cost on a log scale. Filters let users set a minimum score threshold, collapse model variants to show only the best-scoring configuration, and toggle frontier-only view. A lookup table translates the frontier into a budget-based recommendation: find your price ceiling, read off the pick. What the chart reveals is structural. The frontier is thin — only a handful of models at any given moment are non-dominated. The vast majority of available models sit above and to the right of the line, meaning buyers are paying more for less. This is the classic shape of a market where pricing has not caught up to capability convergence: many providers charge premium rates for models that are already beaten by cheaper alternatives. The methodology has real limitations worth flagging. The Intelligence Index is Artificial Analysis's own composite — it is not a community-standard benchmark like MMLU or HumanEval alone, and AA periodically re-bases between versions, meaning cross-snapshot comparisons break. Cached-input discounts, batch pricing, and fast-mode variants are excluded from the blended price, which means the true cost picture for high-volume production workloads could look different. The tool is transparent about this, which counts for something. For buyers, the tool is genuinely useful as a first filter. It collapses a noisy, fast-moving market into a decision framework: if you are not on the frontier, you are overpaying. The daily refresh via GitHub Actions means the data stays current without manual intervention, and the open-source codebase means anyone can audit or fork the methodology. The broader signal here is that LLM pricing is compressing toward commodity dynamics faster than most vendor narratives acknowledge. When a free dashboard can show that half the market's models are dominated, the pricing power of non-frontier providers evaporates. The winners are buyers who treat model selection as a continuous optimization problem rather than a one-time vendor decision. This is a tool, not journalism — it makes no editorial claims about which model is best in absolute terms. Its value is structural: it forces the question of price-performance into a visual format where dominated models cannot hide behind marketing copy or cherry-picked benchmark results.
