sk_live_. A key that already begins with sk_prem_ keeps that prefix and keeps working. Once the account’s plan changes, the plan decides access, not the prefix.
A Premium key asking for
granularity=hourly or auto is served at daily, not refused. Read the granularity and access fields on the response. /api/gpu/{gpuName}/hourly returns the latest hourly print to a Premium key; its day=true form (the trailing 24 hourly prints), /api/gpu/{gpuName}/realtime, and the current-value /api/gpu/{gpuName}/rolling-24h-average return 403. Daily history at /api/gpu/{gpuName}/rolling-24h-average/history remains Premium. GET /api/otpi/hourly follows the Token Price Index window, including Index and Premium.backend/tests/data-dictionary-gating.test.js).
What each plan can read
Power Markets is a Premium-included analytics dataset served on the REST API at
GET /api/power/markets (daily wholesale prices, traded volume, and system load, history from 2017-01-03) and rendered on the platform charts at data.ornn.com/analytics/power. The hourly grid-operations routes — GET /api/power/load for actual and forecast demand, and GET /api/power/fuel-mix for generation by fuel — are Full-only. None of these routes has a free/Index tier. Every other Premium-included analytics dataset has a REST route too: the Token Volume Index at GET /api/token-volume, Model Frontier at GET /api/model-frontier, Compute Buyers at GET /api/compute-buyers, and the forward curve at GET /api/forward (five curves on Premium; B300, GB200, and GB300 require Full), alongside the existing Workload Cost (GET /api/workload) and LLM Coding (GET /api/llm-coding/*) routes. The Token Volume Index page covers all 11 OTPI labs (selectable share, a shared growth baseline, and an open/closed split); the REST route publishes the closed-lab series. See The Token Volume Index. On MCP, Token Volume has its own tool; Model Frontier and Compute Buyers are REST-only (Model Frontier and Compute Buyers because their full-snapshot payloads exceed the MCP result-size cap).
Public GPU window (Index)
H100 SXM · H200 · B200 · A100 SXM4 · RTX PRO 6000 WS
Public OTPI window (Index)
anthropic · openai · google · deepseek
The other seven labs — minimax, xiaomi, qwen, moonshotai, z-ai, mistralai, meta-llama — are Premium and Full only.
The Closed Source and Open Source Token Indexes, which composite all 11 labs into one price per weight class, are on the same trailing 3-month window without a key at GET /api/otpi/weight-class; a key returns the complete history. The Blended Token Index is Premium and Full only.
API keys
Keys are created in Settings → API Keys. One key follows the account. A plan change rewrites the plan stored on that key, so the same secret sees the new plan. Upgrade from Premium to Full, or a downgrade to Premium or Index, does not require a new key and does not revoke the old one. Manual revoke and delete still deactivate a key. Every new key is printedsk_live_, on Trial, Premium, and Full. A key that already begins with sk_prem_ keeps that prefix and keeps working. A key that has not been through a plan change since this rule shipped keeps the older rule: a sk_prem_ key, or one minted on Premium, is Premium; any other is Full.
Keys a member mints through their organization’s access are revoked when that access ends, or when they leave the organization.
See Manage API keys and Authentication.
Payments
- Index — no invoice.
- Trial — no invoice. 72 hours, starting on first visit. The Premium surface with a clock. Request from data.ornn.com/pricing or data@ornn.com.
- Premium — $500 per month, server-pinned. Self-serve: Settings → Billing, or the subscribe button on
/pricing. A saved card is charged automatically; otherwise Stripe emails a hosted invoice. Cancel by writing to data@ornn.com. Bound by the Premium Data Use Addendum. - Full — custom, organization-grained, invoiced directly. Talk to data@ornn.com. Operators can also grant a timed Full trial (same Full surface, with an end date) to one person (
set_access) or the whole organization (set_org_access). When that clock ends, product access and API keys end with it. A teammate approved while the organization is on a timed Full trial receives that same end date, not a permanent seat.
MCP
Access follows the same ladder. Paid tools need an own grant: a Full organization floor covers the product, not the unscoped MCP service key.
See Ornn MCP.
Rate limits
Public Index and catalog routes that use the public limiter are 60 requests / minute / IP. Sending a key does not replace that limiter.POST /api/chat is the exception: 20 messages / minute / key, because each call spends model tokens. See Rate limits.
