This market resolves to the company that authored the single most-used model on OpenRouter during December 2026, measured by token volume and computed from OpenRouter's official rankings dataset.

Determination. Using OpenRouter's daily rankings dataset (GET /api/v1/datasets/rankings-daily), the determining query is start_date=2026-12-01 and end_date=2026-12-31 (UTC, inclusive). For each model (model_permaslug), total_tokens (= prompt_tokens + completion_tokens) is summed across all returned days in that window. The model with the highest December total is the "#1 model," and the market resolves to that model's author company. Tokens are not summed across a company's models — only the single highest-total model determines the outcome.

Author attribution. The author company is the prefix of the model_permaslug before the slash (e.g., anthropic/claude-opus-4.8 → Anthropic; xiaomi/mimo-v2.5 → Xiaomi), mapped to the options above, not the inference provider that routes the traffic. OpenAI and Microsoft count as separate companies. A third-party fine-tune is attributed to whatever author prefix OpenRouter assigns its listing, not the base-model creator. If the author company has been acquired, renamed, or split during 2026, the model is attributed to the company controlling it as of December 31, 2026.

Model = permaslug. Ranking is by model_permaslug, which is the unit OpenRouter reports. Distinct versions or variants listed under distinct permaslugs are distinct models and are not merged; free and paid usage that OpenRouter reports under a single permaslug is counted together under that permaslug.

The other row. Each day's dataset lists the top 50 models plus one aggregated other row covering all models outside the daily top 50. The other row is not a model and is ignored for determining #1; it is never attributable to any company.

Cloaked / stealth models. If the #1 model is cloaked — listed under a placeholder author such as openrouter/... (e.g., a stealth model like "Owl Alpha") — it is attributed to whichever real company OpenRouter has publicly confirmed as its author by the resolution date. If its author remains a placeholder or is officially undisclosed at resolution, the market resolves to "Other."

Traffic included. All token volume in the dataset counts, including free-tier and promotional traffic, since it is not separately identifiable.

Source of truth (in order of priority):

1. Primary — OpenRouter's official rankings-daily dataset, queried as above. This is the reproducible source of record.

2. Fallback A — if that endpoint is unavailable, renamed, or returns no usable December data, the #1 model is read from the "Top this month" view of OpenRouter's "Top Models" leaderboard (openrouter.ai/rankings, overall list, no category filter) as displayed at 23:59 UTC on December 31, 2026.

3. Fallback B — if both the dataset and that page are gone or restructured, this market follows whichever public usage ranking OpenRouter itself designates or links as the successor (via openrouter.ai, its documentation, or official announcements), using its closest equivalent to a most-used-model-by-tokens measure for the end of 2026.

4. N/A — resolves N/A only if no OpenRouter-published usage dataset or ranking exists in any form at resolution.

Methodology. This market follows OpenRouter's usage-volume ranking as it exists at resolution. If OpenRouter changes how token totals or rankings are computed, the updated figures govern rather than any prior methodology — provided the ranking still measures individual models by token-usage volume. If OpenRouter's reported figures cease to be a usage-volume measure (for example, shift to revenue or a quality benchmark), resolution falls to the next available source above that still reflects token usage, preserving the original intent of "the most-used model by tokens." Token counts are taken as OpenRouter reports them, including that counts come from each provider's own tokenizer and are not perfectly comparable across models — the same basis as OpenRouter's public chart.

Resolution timing. Resolves in January 2027, once the dataset's December 2026 rows are complete and stable (i.e., end_date=2026-12-31 is fully populated and no longer changing). Revisions published after that point do not reopen the market.

"Other." Resolves to "Other" if the #1 model's author is not one of options 1–11, or if that author is a placeholder / officially undisclosed at resolution.

Judgment. Where these criteria are unambiguous, the criteria govern even if the result is surprising. Where they are ambiguous or contradictory, the market resolves to its spirit — "which company made the single most-used model on OpenRouter in December 2026, by token throughput" — as conveyed to traders here. Ambiguities will be fixed promptly once flagged; betting through an un-flagged ambiguity is at the bettor's own risk.