> ## Documentation Index > Fetch the complete documentation index at: https://data.ornn.com/docs/llms.txt > Use this file to discover all available pages before exploring further. # Data dictionary > Every dataset and field a subscriber receives, with Basic, Premium, and Full coverage. Version 4.0. Every dataset and every field a subscriber receives, tagged by plan. Delivered through [data.ornn.com](https://data.ornn.com) and the REST API at `https://api.ornnai.com`, with CSV/XLSX export from charts and tables on **Full** only. This is version **4.0**, effective August 19, 2026. It supersedes the v3 dictionary (Public vs Full Data only) by adding **Premium** (\$500/month, daily grain) as a first-class column. Compare plans at [data.ornn.com/pricing](https://data.ornn.com/pricing). Grain and key rules are in [Access tiers](/docs/access-tiers). When those rules change, this page must change with them — `backend/tests/data-dictionary-gating.test.js` fails the PR otherwise. Ornn Data LLC, 447 Broadway FL 4, New York, NY 10013. `access` on a keyed response is `premium` or `full`. On the Basic window it is `public-3mo` (GPU) or `public-1mo` (OTPI). ## 1. Datasets | # | Dataset | What it measures | Unit | Cadence | Basic | Premium / Trial | Full | | -- | -------------------------- | ------------------------------------------------------------------ | --------------------------------------------------- | ------------------------------- | ------------------------------------- | ------------------------------------ | -------------------------------- | | 2 | Compute Price Index (OCPI) | Cost to rent GPU compute | USD per GPU-hour | Hourly prints, daily settlement | 5 of 6 GPUs, trailing 3 months, daily | All 6 GPUs, all-time, **daily only** | All 6 GPUs, all grains, all-time | | 3 | Compute analytics | Volatility and utilization on the GPU index | Decimal fraction | Hourly or daily | None | Daily, all-time | All grains, all-time | | 4 | Forward Curves | Forward marks by tenor for each tracked GPU | USD per GPU-hour | Daily marks | Published types only | Live curve | Live curve | | 5 | Memory Price Index | Component and module spot prices | USD per unit | Each weekday | Catalog only | All 14 types, full history | Same | | 6 | Token Price Index (OTPI) | Lab-level blended token price | USD per million tokens | Daily settlement | 4 of 11 labs, trailing 1 month | All 11 labs, all-time | Same | | 7 | Token Volume Index | Daily token volume and share by lab | Index, base 100 | Daily settlement | None | Included | Included | | 8 | Workload Cost Index | Cost per request, mix-purged unit cost, workload intensity | Cents per request, index points, tokens per request | Daily settlement | None | Included | Included | | 9 | Model Frontier | Benchmark capability against price per task | Score against USD per task | On benchmark refresh | None | Included | Included | | 10 | LLM Coding Analytics | Public-GitHub PRs, merges, commits by coding tool | Counts and percent | Daily | Product catalog only | Included | Included | | 11 | Power Markets | Wholesale US power prices, volume, and load | USD per MWh, GWh | Daily, history from 2017-01-03 | None | Included | Included | | 12 | Datacenter Sites | Tracked AI datacenter sites, operators, power, hardware, timelines | Mixed | Curated | Coordinates only | Coordinates only | Full payload and API | | 13 | Compute Buyers | Funded AI-native startups classified by workload | USD raised | Snapshot | None | Included | Included | Premium adds RTX PRO 6000 WS, the seven OTPI labs outside the Basic four, unlimited **daily** history on both windowed series, and sections 3–11 and 13. Full adds hourly and real-time grains, CSV/XLSX export, and the full datacenter-site payload. Trial is Premium with a 72-hour clock. ## 2. Compute Price Index (OCPI) Per-GPU price series calculated from anonymized executed transactions. The index reflects transactions, not list prices. Published hourly on the hour and aggregated to a daily average. One region is populated, global. **Tracked GPUs (6).** RTX PRO 6000 WS is Premium/Full only; the other five are on Basic at a trailing 3 months, daily. | GPU | Basic | Premium | Full | | --------------- | --------------- | --------------- | -------------------- | | H100 SXM | 3 months, daily | All-time, daily | All grains, all-time | | H200 | 3 months, daily | All-time, daily | All grains, all-time | | A100 SXM4 | 3 months, daily | All-time, daily | All grains, all-time | | B200 | 3 months, daily | All-time, daily | All grains, all-time | | RTX 5090 | 3 months, daily | All-time, daily | All grains, all-time | | RTX PRO 6000 WS | None | All-time, daily | All grains, all-time | | Field | Type | Unit | Description | Example | Access | | --------------------------- | ---------------- | ---------------- | --------------------------------------------------- | ------------------------ | ----------------------- | | `gpu_type` / `gpu_name` | string | none | The GPU model the series is for | H100 SXM | All | | `region` | string | none | Geographic market, empty or `global` | global | All | | `date` | string, date | none | Settlement date of a daily index value | 2026-07-17 | All | | `timestamp` / `recorded_at` | string, ISO 8601 | UTC | Date and time of a print or history point | 2026-07-17T20:00:00.000Z | All | | `index_value` | number | USD per GPU-hour | Published index value, rounded to a tenth of a cent | 2.00 | All (windowed on Basic) | | `average_usd_per_hour` | number | USD per GPU-hour | Rolling 24-hour average price, published hourly | 2.00 | Full (hourly path) | | `last_updated` | string, ISO 8601 | UTC | When the value was computed | 2026-07-17T20:05:02.000Z | All | | `access` | string | none | Tier served: `public-3mo`, `premium`, or `full` | premium | All | | `granularity` | string | none | Grain actually served: `daily` or `hourly` | daily | Keyed | ## 3. Compute analytics Volatility and utilization derived from the GPU index history, per GPU. No Basic window. | Field | Type | Unit | Description | Example | Access | | -------------------- | ---------------- | --------------------------- | ------------------------------------------------------------------------------------------------- | ------------------------ | -------------------------------- | | `recorded_at` | string, ISO 8601 | UTC | Point timestamp | 2026-07-17T00:00:00.000Z | Premium+, daily; Full, any grain | | `gpu_type` | string | none | GPU model | H100 SXM | Premium+ | | `region` | string | none | Market | global | Premium+ | | `index_value` | number | USD per GPU-hour | Index value at that point | 2.00 | Premium+ | | `rolling_volatility` | number | annualized decimal fraction | Standard deviation of log returns over the rolling window, annualized. 0.80 is roughly 80 percent | 0.20 | Premium+ | | `window_days` | number | days | Lookback applied. Default 30, range 1 to 365 | 30 | Premium+ | Changes are in percentage points, not percent change: both ends are already percentages. | Field | Type | Unit | Description | Example | Access | | ------------------- | ---------------- | ----------------- | -------------------------------------------------------------------- | ------------------------ | -------- | | `recorded_at` | string, ISO 8601 | UTC | Point timestamp | 2026-07-17T00:00:00.000Z | Premium+ | | `utilization_ratio` | number | decimal fraction | Share of tracked capacity currently rented. 0.70 is about 70 percent | 0.70 | Premium+ | | `latestPct` | number, nullable | percent | Most recent day's mean utilization | 70.4 | Premium+ | | `dayPoints` | number, nullable | percentage points | Change against the prior day | 1.2 | Premium+ | | `weekPoints` | number, nullable | percentage points | Change against 7 days prior | -0.8 | Premium+ | | `monthPoints` | number, nullable | percentage points | Change against 30 days prior | 4.1 | Premium+ | ## 4. Forward Curves Forward marks by tenor for each tracked GPU, struck at the daily settlement. No Basic marks API. **Tracked curves (7).** H100 SXM, H200, A100 SXM4, RTX 5090, B200, RTX PRO 6000 WS, B300. **Tenors (5).** 1M (1 month), 6M (6), 1Y (12), 3Y (36), 5Y (60). Curves read on a term basis (average cost to the tenor) or a forward basis (implied rate between two tenors), with the depreciation each curve implies. | Field | Type | Unit | Description | Example | Access | | ---------- | ---------------- | ---------------- | ----------------------------------------------------------- | ------------------------ | -------- | | `name` | string | none | The GPU the curve is for | H100 SXM | Premium+ | | `label` | string | none | Tenor label | 1Y | Premium+ | | `months` | number | months | Tenor in months | 12 | Premium+ | | `price` | number | USD per GPU-hour | The forward mark at that tenor | 1.98 | Premium+ | | `has_mark` | boolean | none | Whether price is a real quote. `false` is a gap, never 0.00 | true | Premium+ | | `asOf` | string, ISO 8601 | UTC | Timestamp of the mark set | 2026-08-12T00:00:00.000Z | Premium+ | ## 5. Memory Price Index Daily spot prices for 14 memory types in three categories, refreshed each weekday. No Basic window on prices; the type catalog is public. **Tracked memory types** | Category | Memory types | | -------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | dram | DDR3 4Gb (512Mx8) 1600/1866, DDR4 8Gb (1Gx8) 3200, DDR4 8Gb (1Gx8) eTT, DDR4 16Gb (2Gx8) 3200, DDR4 16Gb (2Gx8) eTT, DDR5 16Gb (2Gx8) 4800/5600, DDR5 16Gb (2Gx8) eTT | | flash | MLC 32Gb (4Gx8), MLC 64Gb (8Gx8), SLC 1Gb (128Mx8), SLC 2Gb (256Mx8) | | module | DDR4 UDIMM 16GB 3200, DDR5 RDIMM 32GB 4800/5600, DDR5 UDIMM 16GB 4800/5600 | | Field | Type | Unit | Description | Example | Access | | ----------------- | ---------------- | ------- | -------------------------------------------- | ------------------------ | -------------------------------- | | `memoryType` | string | none | The tracked memory product | DDR4 16Gb (2Gx8) 3200 | Catalog: Basic. Prices: Premium+ | | `category` | string | none | `dram`, `flash`, or `module` | dram | Catalog: Basic. Prices: Premium+ | | `price` | number | USD | Latest daily spot price, the session average | 40.00 | Premium+ | | `changePct` | number | percent | Change from the prior reading | 0.00 | Premium+ | | `weeklyHigh` | number | USD | High over the trailing week | 50.00 | Premium+ | | `weeklyLow` | number | USD | Low over the trailing week | 30.00 | Premium+ | | `date` | string, ISO 8601 | UTC | Price date | 2026-07-17T00:00:00.000Z | Premium+ | | `timestamp` | string, ISO 8601 | UTC | Point timestamp on history rows | 2026-07-17T00:00:00.000Z | Premium+ | | `sourceUpdatedAt` | string, ISO 8601 | UTC | Upstream publication time | 2026-07-17T00:00:00.000Z | Premium+ | | `updatedAt` | string, ISO 8601 | UTC | When Ornn ingested the value | 2026-07-17T14:00:00.000Z | Premium+ | ## 6. Token Price Index (OTPI) A daily, volume-weighted blended price per million tokens for each lab, across its paid production models. Free tiers and open-weight self-hosting are excluded. **Tracked labs (11).** The first four are Basic at a trailing 1 month; the rest are Premium/Full only. | Lab | Label | Basic | Premium / Full | | ---------- | ----------- | ------- | -------------- | | anthropic | Anthropic | 1 month | All-time | | openai | OpenAI | 1 month | All-time | | google | Google | 1 month | All-time | | deepseek | DeepSeek | 1 month | All-time | | minimax | MiniMax | None | All-time | | xiaomi | Xiaomi | None | All-time | | qwen | Qwen | None | All-time | | moonshotai | Moonshot AI | None | All-time | | z-ai | Z.ai | None | All-time | | mistralai | Mistral AI | None | All-time | | meta-llama | Meta Llama | None | All-time | | Field | Type | Unit | Description | Example | Access | | -------------- | ---------------- | ---------------------- | ------------------------------------------------ | ------------------------ | -------------------------------- | | `date` | string, date | none | Settlement date the value covers | 2026-07-17 | All (windowed on Basic) | | `lab` | string | none | Model provider identifier | anthropic | All (lab list windowed on Basic) | | `indexPerMtok` | number | USD per million tokens | Blended price per million tokens, full precision | 1.234567 | All (windowed on Basic) | | `computedAt` | string, ISO 8601 | UTC | When the value was computed | 2026-07-18T12:00:00.000Z | All | | `access` | string | none | Tier served, `public-1mo`, `premium`, or `full` | premium | All | ## 7. Token Volume Index Daily token volume per lab, indexed to 100 at that lab's first settled day, with each lab's share of the closed-lab total. Daily granularity. No Basic window. | Field | Type | Unit | Description | Access | | ------------------ | ---------------- | ---------------- | ---------------------------------------------------------- | -------- | | `date` | string, date | none | Settlement day | Premium+ | | `lab` | string | none | Model provider identifier | Premium+ | | `totalTokens` | number | tokens | Settled daily token volume for that lab | Premium+ | | `growthIndex` | number | index points | That lab's volume indexed to 100 at its baseline day | Premium+ | | `share` | number | decimal fraction | The lab's share of the closed-lab daily total | Premium+ | | `totalGrowthIndex` | number, nullable | index points | The all-lab total, indexed to 100 at the total baseline | Premium+ | | `baselines` | array | mixed | Per-lab `{ lab, date, totalTokens }` the index is based at | Premium+ | | `totalBaseline` | object, nullable | mixed | `{ date, totalTokens }` for the all-lab total | Premium+ | ## 8. Workload Cost Index The OTPI companion that prices the request rather than the token, for the three closed labs. History from 2026-06-11. No Basic window. **Tracked providers and model classes.** Classes are exact partitions of a provider's daily rows, so class rows sum back to the aggregate. | Series | What it measures | Unit | | ------------------- | -------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------- | | Cost per Request | Settled spend over paid calls for the node's model mix that day. Includes mix, so it moves when users shift between models | cents per request, cents per image for image classes | | Like-for-Like Index | Mix-purged unit cost. Separates true price and efficiency change from savings that come from routing to cheaper models | index points, base 100 | | Workload Intensity | How heavy a typical call is: tokens served per request across the node's mix | tokens per request | | Provider | Aggregate | Model classes | | --------- | --------- | --------------------------------------- | | anthropic | yes | opus, sonnet, haiku, fable | | openai | yes | flagship, mini, nano, pro, codex, image | | google | yes | flash, pro, lite, image | | Field | Type | Unit | Description | Example | Access | | ----------------- | ---------------- | ------------------------------- | ----------------------------------------------------- | ------------------------ | -------- | | `date` | string, date | none | Settlement date | 2026-07-13 | Premium+ | | `lab` | string | none | anthropic, openai, or google | anthropic | Premium+ | | `family` | string | none | `all` for the provider aggregate, or a model class | all | Premium+ | | `cCents` | number | cents per request | Cost per Request | 4.82 | Premium+ | | `lfl` | number | index points | Like-for-Like Index | 96.4 | Premium+ | | `qKtok` | number | thousands of tokens per request | Workload Intensity | 1.82 | Premium+ | | `priceEffect` | number | percent contribution | Same-model unit-cost contribution to the day's change | 0.4 | Premium+ | | `intensityEffect` | number | percent contribution | Tokens-per-request contribution | -1.1 | Premium+ | | `mixEffect` | number, nullable | percent contribution | Model-mix contribution | 0.2 | Premium+ | | `computedAt` | string, ISO 8601 | UTC | When the row was computed | 2026-07-14T00:00:00.000Z | Premium+ | ## 9. Model Frontier Benchmark capability against price per task, for every model and thinking level. Cost is priced off the index: the lab's OTPI dollars per million tokens multiplied by the model's tokens per task. No Basic window. Capability axes: Overall (Epoch ECI), Coding (SWE-bench Verified), Reasoning (GPQA Diamond), Math (AIME 2024/25). A model with no real score on an axis is dropped from that view, never estimated. | Field | Type | Unit | Description | Access | | ----------------- | ---------------- | ---------------------- | ------------------------------------------------------------------------------ | -------- | | `slug` / `name` | string | none | Model identifier and display name | Premium+ | | `lab` | string | none | The lab that serves the model | Premium+ | | `cap` | number | ECI index points | Epoch Capability Index composite score | Premium+ | | `tpt` | number | tokens | Tokens per task | Premium+ | | `price` | number | USD per million tokens | Per-model price used for the cost axis | Premium+ | | `pIn` / `pOut` | number | USD per million tokens | Input and output prices | Premium+ | | `totalTokens` | number | tokens | Volume behind the model | Premium+ | | `costPerTask` | number | USD per task | price × tpt / 1e6 | Premium+ | | `bench.coding` | number | percent | SWE-bench Verified | Premium+ | | `bench.reasoning` | number | percent | GPQA Diamond | Premium+ | | `bench.math` | number | percent | AIME 2024/25 | Premium+ | | `model` | string | none | Model identifier (thinking-level rows) | Premium+ | | `effort` | string | none | minimal, low, medium, high, xhigh, max, thinking, or default | Premium+ | | `rank` | number | 0 to 5 | Ordering from least to most thinking | Premium+ | | `cost` | number | USD per task | OTPI-rebased where the lab has an OTPI price, else the measured provider price | Premium+ | | `tokensEst` | number, nullable | tokens per task | Output tokens per task | Premium+ | | `timeSec` | number, nullable | seconds | Time to solve | Premium+ | | `speedTps` | number, nullable | tokens per second | Output speed used for timeSec | Premium+ | | `speedMeasured` | boolean | none | true if measured, false if a median fallback | Premium+ | | `costBasis` | string | none | `otpi` or `measured` | Premium+ | **Tracked benchmarks (11).** WeirdML, Aider Polyglot, CursorBench, DeepSWE, ALE-Bench, ARC-AGI, ARC-AGI-2, ProofBench, Surface Evolver, CritPt, The Agent Company. ## 10. LLM Coding Analytics Daily public-GitHub activity for the major AI coding tools. No Basic window on values; the product catalog is public. **Tracked tools (5).** claude (Claude Code), codex, copilot (GitHub Copilot), cursor, devin (Cognition). PRs and merge rates for all five; commits for claude only. **Products (13).** A tool crossed with a metric. Each has an `id`, `label`, `category`, `kind` (count or pct), `compare` flag, and `seriesLabels`. Public-GitHub signals only. Private repos, self-hosted instances, and IDE-side telemetry are out of scope. | Field | Type | Unit | Description | Access | | ------------------ | ---------------- | -------------------- | ----------------------------------------------------------------------- | -------------------------------- | | `date` | string, date | UTC | The day the activity was observed | Premium+ | | `tool` | string | none | One of the five tools | Catalog: Basic. Values: Premium+ | | `commits` | number | count | Public-GitHub commits attributed to the tool that day, claude only | Premium+ | | `prs` | number | count | Pull requests opened from the tool's branch namespace | Premium+ | | `merged` | number | count | Of those PRs, how many have merged so far | Premium+ | | `merge_rate` | number | percent, one decimal | 100 × merged / prs | Premium+ | | `value` / `value2` | number | count or percent | The series value; value2 carries the second series on a compare product | Premium+ | | `changePct` | number, nullable | percent | Day-over-day change | Premium+ | | `updatedAt` | string, ISO 8601 | UTC | When the value was written | Premium+ | ## 11. Power Markets Daily wholesale power prices across every US market, with history from 2017-01-03. Six markets are ICE exchange trades carrying the volume behind each print; ERCOT is its own day-ahead clearing price, shown against grid load. No Basic window. Traded volume and system load are separate fields and never share an axis: a clearing price has no traded volume, so ERCOT carries the energy the grid actually moved. Daily peak-block prices per hub, not hourly nodal. **Tracked markets** | Market | Depth series | Hub(s) | Source | | -------------- | -------------------------------- | ------------------------------------------------ | ------ | | WECC Northwest | Traded volume | Mid C Peak | ICE | | PJM | Traded volume | PJM WH Real Time Peak | ICE | | CAISO | Traded volume | NP15 EZ Gen DA LMP Peak, SP15 EZ Gen DA LMP Peak | ICE | | ISO-NE | Traded volume | Nepool MH DA LMP Peak | ICE | | WECC Southwest | Traded volume | Palo Verde Peak | ICE | | MISO | Traded volume | Indiana Hub RT Peak | ICE | | ERCOT | System load, plus off-peak price | HB\_NORTH | ERCOT | | Field | Type | Unit | Description | Access | | ----- | ---------------- | ----------- | ---------------------------------------------------------------- | -------- | | `d` | string, date | none | Trade or clearing date | Premium+ | | `p` | number | USD per MWh | Peak-block price, the headline series | Premium+ | | `o` | number, optional | USD per MWh | Off-peak price, hours ending 1 to 6 and 23 to 24 of the same day | Premium+ | | `t` | number, optional | GWh | Traded volume, exchange-traded markets only | Premium+ | | `l` | number, optional | GWh | System load delivered, clearing markets only | Premium+ | | `g` | number, optional | count | Trade count behind the print | Premium+ | ## 12. Datacenter Sites Every tracked AI datacenter site worldwide, with the operator, power capacity, hardware, and build timeline behind each one. **Full only** for the payload. Basic and Premium see coordinates on the public map thumbnail. Sites can be rolled up by country or by GPU type into `site_count`, `total_power_mw`, and `total_gpu_qty`. | Field | Type | Description | Access | | ---------------------------------------------- | ---------- | ------------------------------------------------------------- | ----------------------------------------- | | `company` | string | The operator running the site | Full | | `siteName` | string | The named facility or campus | Full | | `streetAddress` / `city` / `state` / `country` | string | Physical location | Full | | `latitude` / `longitude` | number | Coordinates | Basic+ (points view); Full (with payload) | | `dcTier` | string | Datacenter tier or classification, for example Tier III+ | Full | | `ownershipModel` | string | Owned, leased, or colocation | Full | | `coloPartner` | string | Colocation partner where the site is hosted rather than owned | Full | | `gpuTypes` | string | GPU models deployed or planned | Full | | `gpuQtyKnown` | number | Confirmed GPU count where reported | Full | | `gpuQtyEst` | number | Estimated GPU count where not reported | Full | | `chipGeneration` | string | Hardware generation deployed or committed | Full | | `powerCapacity` | number, MW | The site's power envelope | Full | | `category` | string | Business category, for example GPU-as-a-Service | Full | | `isNeocloud` | boolean | Whether the site is a neo-cloud (GPU-as-a-service) operator | Full | | `status` | string | Lifecycle stage: announced, under construction, operational | Full | | `leaseTerm` | string | Lease length where the site is leased | Full | | `keyCustomer` | string | Anchor tenant or key customer where known | Full | | `contractAnnounced` | date | Contract announcement | Full | | `gpuOrderDate` | date | GPU order placed | Full | | `constructionStart` | date | Construction start | Full | | `phase1LiveDate` | date | First phase live | Full | | `fullCapacityTarget` | date | Full-capacity target | Full | | `sourceNotes` | string | Provenance and caveats for the row | Full | ## 13. Compute Buyers AI-native startups that raised 10 million dollars or more in a single event, classified by the workload their compute is for. Built from SEC Form D filings and announced rounds, July 2024 to July 2026. No Basic window. | Field | Type | Unit | Description | Access | | ---------------------- | ---------------- | ---- | ------------------------------------------------------------------------------ | -------- | | `name` / `displayName` | string | none | Company legal name and display name | Premium+ | | `tier` | string | none | T1 trains models, T2 heavy inference or AI infrastructure, T3 GPU-adjacent HPC | Premium+ | | `workload` | string | none | The workload classification | Premium+ | | `forWhat` | string | none | What the compute is for | Premium+ | | `what` | string | none | What the company does | Premium+ | | `sold` | number, nullable | USD | Amount raised in the event | Premium+ | | `date` | string, date | none | Event date | Premium+ | | `location` | string | none | Company location | Premium+ | | `domain` | string, nullable | none | Company domain | Premium+ | | `url` | string | none | Source link for the raise | Premium+ | ## Keys, payments, and grain | | Basic | Premium / Trial | Full | | ----------------- | ----- | ----------------------------------------------------- | --------------------------- | | Payment | Free | **\$500/month** Stripe, or a 72-hour trial | Custom, org-grained, Rillet | | API key | None | `sk_prem_` | `sk_live_` | | Hourly / realtime | No | 403 on dedicated routes; other routes served at daily | Yes | | Neocloud API | No | 403 | Yes | | CSV / XLSX export | No | No | Yes | The key prefix is the tier. See [Access tiers](/docs/access-tiers). When API gating or Premium billing changes, update this page, Access tiers, and regenerate the internal Word file from `frontend/app/internal/data-dictionary/build-docx.py`.