What share do we actually see?
Primary definition: continuous, directly metered tokens in the database divided by a modeled worldwide or U.S. denominator. The disclosure ledger is shown separately because it is broad but stale and asynchronous.
Current world denominator
The v0.3 public tokenomics low / base / high band, using fresh lab disclosures and explicit app, financial, China-residual, and hosted-model assumptions.
Coverage arithmetic
The base estimate uses a smoothed seven-day router average, not a volatile single day.
Usage is compounding much faster than price is falling.
OpenRouter is the direct platform series. The global line is a modeled quarterly band. They answer different questions and are deliberately not spliced into one series.
OpenRouter tokens per day
Exact public-platform tokens; daily and 7-day moving average.
Modeled global tokens per day
Historical quarterly public-model band with the latest point replaced by the v0.3 bottom-up lab/app/financial estimate; trillions per day. Changes reflect both real growth and better disclosure coverage.
Relative strength by model lab
Modeled share of aggregate worldwide token demand, with OpenRouter retained below as a separate high-frequency platform signal. Direct disclosures are pinned first; only the residual pool uses app/router allocation priors.
Aggregate daily tokens by model lab
Mutually exclusive low / base / high allocation of the same global demand denominator. Share ranges can overlap and do not independently sum to 100%; base shares do.
Modeled global lab leaderboard
Momentum uses each lab’s freshest comparable public signal—tokens, ARR, users, or interactions—and is not a like-for-like global token-growth series.
| # | Lab | Low T/day | Base T/day | High T/day | Base share | Share envelope | Momentum proxy | Class |
|---|
Model-family attribution
The most granular defensible global view is model family, not exact version. Production routing among GPT, Claude, Gemini, Doubao, Qwen, Grok, and other versions is not publicly disclosed.
| # | Model family | Lab | Base T/day | Base share | Confidence | Exact version split | As of |
|---|
OpenRouter high-frequency signal
Exact public-platform tokens and momentum. Useful for launches and open-model rotation; not used as a global market-share meter.
Monthly average daily tokens by leading lab—OpenRouter
Top eight labs by current platform volume; all others bucketed as Rest. Monthly averages keep partial August comparable with full months.
Current OpenRouter lab leaderboard
Sorted by latest seven-day platform token volume.
| # | Lab | 7d tokens | Share | WoW | 30d tokens |
|---|
Model strength and indicative list-price economics
Token volume provides adoption momentum; current list price provides a rough revenue-intensity axis. The cost math is not realized billing because prompt/output mix, caching, discounts, and historical prices are unavailable.
Volume versus blended list price
Top 30 models; price assumes 75% input / 25% output and current OpenRouter list prices. Free models appear at zero.
Frontier versus budget cost index
MyTokenTracker fixed baskets, $/million tokens, 3:1 input:output. The public history is short and currently flat.
Current model leaderboard
Indicative cost applies today's blended list price to all seven-day tokens. It is a comparable intensity measure, not actual provider revenue.
| # | Model | Lab | 7d tokens | Share | WoW | Blended $/M | Indicative 7d list cost |
|---|
Where the tokens are going
Publicly attributed applications and sampled weekly task filters provide a directional demand map. Neither is a complete application census.
Top public apps—trailing 30 days
OpenRouter public attribution; trillions of tokens.
Task-filter token estimates
Latest sampled/upsampled week. Filters should not be summed into a market total because labeling may overlap.
Public-data tokenomics model
A bottom-up demand, capacity, and economics model built from public disclosures and reproducible benchmarks. The model exposes its residual instead of using a global multiplier to make the numbers balance.
Demand components and uncertainty
Low / base / high trillion processed tokens per day. Cross-check rows are labeled and excluded from the additive total.
Demand versus public capacity
The public supply bar includes only a dated xAI operating-fleet disclosure. It is a feasibility subset—not global supply.
The residual is primarily an incompleteness diagnostic. Amazon, Google, Microsoft, Meta, CoreWeave, and China now appear in the native-unit evidence ledger, but remain outside the T/day subtotal because public SKU, workload, utilization, or power-boundary data are insufficient. It is not unmet demand and it is not assigned to xAI.
Component ledger
Observed, modeled, and capacity-derived rows stay separate. Additive status determines whether a demand row enters the total.
| Component | Channel | Class / confidence | Low | Base | High | Additive |
|---|
Public supply evidence in native units
Chip counts, MW, capability envelopes, and national FLOPS are preserved without manufacturing a common token conversion. Only rows with a compatible benchmark enter the T/day supply subtotal.
| Operator / ecosystem | Status | Low | Base | High | Native unit | Token conversion | As of |
|---|
Component methodology
Open each component for its exact formula, transformation procedure, source links, and known failure modes.
Assumption ledger
Every direct input and modeled parameter, with source, date, confidence, and rationale.
| Component | Channel | Metric | Low | Base | High | Unit | Confidence | As of |
|---|
Unresolved channels
These are intentionally left visible rather than filled by an opaque market multiplier.
Definitions, assumptions, and sources
The dashboard separates metered router traffic, modeled global demand, disclosure representation, sampled classification, and cost proxies.
Capture-rate definitions
- Worldwide direct capture: latest seven-day average OpenRouter public-platform tokens divided by the v0.3 global low/base/high demand band.
- U.S. direct capture: a modeled OpenRouter U.S. numerator divided by modeled ex-China U.S. demand. Base case applies 47% to both; the range allows router geography and market geography to differ.
- Disclosure representation: additive current direct-token rows divided by the v0.3 global band. Overlapping country, product, API, and router rows are excluded.
Key assumptions
- The current global band comes from the versioned v0.3 lab/app/financial partition rather than carrying forward one historical anchor.
- The U.S. bridge subtracts the current 220T/day base China envelope before applying a 47% base U.S. share. This remains low confidence.
- OpenRouter token definitions and company disclosures are accepted as reported; tokenizer, multimodal, and hidden-reasoning treatment are not harmonized.
- Current model prices are blended only for indicative economics. They are not used as realized billing except where an explicit financial backsolve says so.
Source ledger
What is missing
Continuous first-party Claude and many enterprise API meters; AWS Bedrock, Azure, and Vertex reseller totals; exact production routing by model version; private/ZDR gateways; self-hosted and on-device inference; Meta recommendation/ads inference; multimodal token equivalents; and reliable end-user geography. These omissions remain explicit rather than being filled by a single opaque multiplier.