Skip to content

Reports

Living reports: always-current reference data on the LLM market, updated on a fixed cadence and versioned — every edition is archived, so you can see how the ecosystem moves, not just where it stands. For analysis and commentary, see the blog.

The decision guide, three steps: pick your weight class, pick open vs closed within it, pick the model — with current benchmarks, prices and the strategic trade-offs (deprecation, control, licenses) sourced per claim. Updated monthly.

Which models are actually worth their price: every current flagship — open and closed — on output price vs the Artificial Analysis Intelligence Index, with the Pareto frontier computed from the data. Updated monthly. Latest edition: 2026-10 (updated 2026-11-05) — Kimi K3.1 Pro (55) flips to open weights, narrowing the open-vs-closed gap to 3 points; Claude Opus 5.5 remains the index leader.

The open-weight models we serve, and why: licenses with their real fine print, parameters, context windows, current per-token prices and benchmark scores. Numbers refreshed monthly; roster updated when the lineup changes.

At what monthly token volume a flat unlimited block beats per-token pricing, for every model we serve, with the savings factor at realistic volumes and the honest ceiling a single key will not go past. Recomputed from live prices on every build.

What a million tokens costs, tracked month over month — open vs closed, with price trajectories that draw themselves as editions accumulate, and the metric that matters: dollars per intelligence point. Updated monthly.


The underlying data for each edition is versioned; if you want it in machine-readable form, tell us.