Funding & independence

Age of LLM accepts API credits and compute from model providers, inference providers and platforms. It does not accept payment for placement, coverage, or favourable treatment.

Every benchmark that takes money from the labs it ranks has to answer the same question, so this page answers it before anyone asks. The rules below are binding, they apply to every contributor without exception, and every contribution received is listed at the bottom of this page.

The commitments

• Sponsors fund compute, never methodology. No funder has any input into the rules, the system prompt, the scoring, the tie-breakers, the map, or which models enter the corpus.
• No preview. No result is shared with any party before publication, including with the funder of the model concerned. Nobody sees a score early, and nobody gets to ask for a rerun after seeing one.
• Identical protocol. A funded model is tested exactly as any other: same engine version, same prompt, same reasoning effort, same opponent (the reigning champion), same two matches.
• Full disclosure. Every contribution is listed below with its amount and what it paid for, before the results it funded are published.
• Right to publish anything. Results are published regardless of how a funded model performs, and so are our own methodological failures. See the side-bias audit below.

What a fair test costs

Since 5 September 2026 the board is a throne. There is one reigning champion, and a new model plays it and nobody else: two matches, one from each side of the map. Win the tie and the crown changes hands. Lose it and nothing on the board moves. What it costs to test a model no longer depends on how good the model turns out to be.

Every figure below is the amount the provider actually billed, read back out of each replay, not a price × token estimate. The distinction is not cosmetic: prompt caching made one model 22 % cheaper than its estimate in the opening, so the estimate was a number we knew to be wrong.

Half of every bill is the champion's, so these figures move with the crown. The current champion, Gemini 3.8 Flash, is one of the cheaper models on the board and is billed at a 50 % launch discount today. A more expensive champion makes every challenge cost more.

What independence has already cost us

In August 2026 an internal check found that the player-2 slot had won 71.7% of decisive matches in our own corpus, which is the kind of number that quietly invalidates a leaderboard. Rather than leave it, we ran a 57-match controlled mirror experiment: the same model on both sides, so model strength is held constant by construction and any remaining imbalance can only come from the engine.

The engine turned out to be neutral: player 2 won 51.9% of decisive mirror matches (Wilson 95% CI [38.9–64.6], exact binomial p = 0.89), significantly different from the corpus figure (Fisher exact p = 0.047). The imbalance came from our own match scheduling, not from the game, so no published result needed correcting and every A-vs-B match had been a fair fight. Matches are now side-swapped by default.

August 2026, again: the engine was penalising the models for its own defects. Before the first match of the current corpus we re-read all 54 archived replays and asked what had never been asked of them: of the 300 actions the engine rejected, how many could the model actually have avoided? Three answers came back and all three were ours. Every one of the 60 "unit not found" rejections, across 19 models, came from the engine assigning a new unit's id only after the model had already submitted its turn: unguessable by construction. Aiming a building into unscouted territory answered "there is already a building here", which handed out a free map probe. And bumping into a unit you cannot see, which is how a fog-of-war game reveals its board, was being counted in the published illegal-action column, penalising the very probing the fog exists to require. All three are fixed in game engine 0.17.0, and on the 12 opening matches that followed the illegal-action rate was 2.1 % against 5.3 % on the old corpus.

The side question, and what the matches since have said. Across the 12 opening matches the right-hand side of the map won 10 (exact binomial p = 0.039). We noticed that pattern during the campaign and tested it on the same 12 matches, which is the weakest form of evidence there is, so we reported it as open and kept counting. The 26 decisive challenge matches played since then are a clean test, and the right-hand side won 21 of them (p = 0.002). Over all 38 decisive matches played on game engines 0.17.0 and 0.18.0 it has won 31. The 57-match mirror experiment above found nothing of the kind on an earlier engine, and we have not yet found the cause.

The crown is protected by construction: every challenge is played once from each side, so a side advantage cancels inside the tie. What it changes is how ties end. In 6 of the 14 ties played so far, each model won the match it played from the right and the tie was settled on speed, and 3 of the 5 challenges for the throne ended that way. We are publishing this before we can explain it, because that is the point of this section.

We publish all of this because a benchmark that never reports a problem with itself is not a benchmark that has none. Full method and figures are in the README.

On verifiability

The full replay corpus, the leaderboard data and the complete methodology are public. The engine itself is deliberately not published: releasing it would allow models to be trained against it, which is the one thing a benchmark built on adversarial play cannot afford. It is available for private audit to any organisation that wants to verify a result before relying on it.

Contributions received

Everything else, including all 40 matches played since 15 August 2026, has been paid for personally.

Contributing compute

If you build, host or resell a model and want it tested properly, the ask is API credits or inference, never cash. Contributors are credited as the compute provider on every match run through them, under the rules above.

Reach out on X (@ageofllm) or open an issue on GitHub.