Best LLM for every budget, updated daily
This is a daily-updated dashboard that plots every model in the Artificial Analysis Intelligence Index against its blended API price to find the best LLM for any given budget. Models on the "value frontier" are those where nothing cheaper is also smarter; every other model is dominated, meaning some cheaper point on the line matches or beats it on both price and score.
The main chart shows blended cost per 1M tokens — a 3:1 input:output blend, excluding cached-input discounts, batch pricing, and fast modes — on a log scale against the Intelligence Index, with hover or tab details on each point. By default only the best variant of each model is shown and low scorers are hidden to keep the field readable, but filters let users widen it. The frontier is computed by sorting models by price ascending and keeping every model that scores higher than everything cheaper, with ties on price going to the higher score and ties on score going to the cheaper model.
Other views include a budget lookup table that names the highest-scoring model available at each price point alongside a runner-up, a raw capability ranking that ignores cost entirely, a "What changed" section diffing consecutive daily fetches to flag new, removed, re-scored, or re-priced models, and a sortable all-figures table. That table lists model, maker, intelligence, coding, and math scores, blended/input/output prices per 1M tokens, tokens per second, and time to first token, with names linking to each model's Artificial Analysis page.
Several controls shape the results: the "one row per model" toggle keeps the highest-scoring effort or reasoning variant of each model name (ties going to the cheaper one), and unticking it exposes every variant Artificial Analysis benchmarks separately, such as low, medium, high, xhigh, and max effort. A minimum-score filter hides weaker models from both the chart and the frontier calculation so an old, very cheap model does not anchor the line. Because AA re-bases the Index between versions, users should compare against this page only, not an older snapshot. Data comes from the Artificial Analysis free data API, fetched daily by a GitHub Actions cron, with code and data at github.com/terryds/bestvaluemodel.