Attic Standard Feed

The benchmark, piped into your systems
Structured inference pricing data delivered direct to your systems, 32 queryable fields per record, normalized and verified, refreshed weekly across every vendor we track. Power cost-aware routing, embed rates into your platform, or feed your own analytics.

OVERVIEW
Pricing Data, Not a Dashboard
What Feed is, what it does, why it matters
Stop scraping vendor pricing pages, maintaining custom parsers, and reconciling units across every vendor you buy from. Feed gives you the answer.

Feed is the raw data product. It delivers the complete inference pricing dataset, normalized and structured, directly into your stack. No dashboard, no interface, no abstraction layer, just the data ready to power whatever you build on top of it.

Each record carries 32 fields covering vendor, model, modality, pricing direction, normalized rates, deployment method, region and full audit metadata. Built for cost-aware routing, model selection algorithms and quantitative cost modelling.


THE SCHEMA
32 Fields Per SKU, Normalized
Sample record and field coverage

Feed records follow a stable schema. Every SKU carries vendor identifiers, model metadata, capability flags, normalized pricing, original pricing, regional context and audit fields. Normalized prices use a deterministic conversion formula stored alongside the record, so any figure can be walked back to the rate the vendor published.

GET /v1/sku_indexSample record
{
  "sku_id": "openai_gpt_4o_input_global",
  "vendor_name": "OpenAI",
  "vendor_type": "Model Developer",
  "model_name": "GPT-4o",
  "model_family": "GPT-4",
  "modality": "Text",
  "context_window": 128000,
  "tool_calling": true,
  "open_source": false,
  "direction": "Input",
  "original_price": 2.50,
  "normalized_price": 0.0025,
  "normalized_unit": "USD per 1K tokens",
  "delivery_method": "API",
  "region": "Global",
  "verification_date": "2026-08-03"
}

The full field set groups into six areas.

Group
Fields
Identity
sku_id, sku_plan_name, vendor_id, vendor_name, vendor_type, model_id, model_name
Capability
modality, modality_subtype, model_family, context_window, tool_calling, open_source
Pricing
direction, original_price, normalized_price, normalized_unit, billing_method
Normalization
normalization_method, normalization_formula
Delivery and region
delivery_method, deployment_method, contracting_method, region, country
Provenance
vendor_url, pricing_url, source_url, pricing_notes, verification_date, run_id, null_reasons

CAPABILITIES
Production-Grade Data Infrastructure
What Feed delivers that a scrape does not
Full schema
32 fields per SKU
Vendor identifiers, model capability metadata, normalized and original pricing, deployment context, regional availability and audit metadata. Filter or join on any field.
Normalization
Deterministic cross-vendor comparison
Every price is normalized to a standard unit using deterministic formulas stored alongside the record. Compare like for like across every tracked vendor without writing parsers.
Update frequency
Weekly refresh, audit trail included
Pricing refreshes weekly, and every record carries a verification date confirming when it was last validated against the vendor's published page.
Delivery options
Pull, push, or direct database access
REST API, scheduled CSV or Parquet exports, or read access to a dedicated Postgres replica. Pick the integration pattern that fits your stack.

DELIVERY
Three Integration Patterns
Pull, push, or direct database access

Feed supports three delivery patterns, each suited to a different operational profile. All three carry the same 32-field schema.

Pattern
How it works
REST API
Authenticated endpoints for sku_index, price_index history, index_values, model_registry and vendor_registry, with standard query parameters for filtering and JSON responses.
Scheduled exports
Weekly CSV or Parquet drops to your S3, GCS or Azure Blob bucket. Full snapshot or incremental delta, schema-stable, built for warehouse ingestion.
Direct database access
A read-only Postgres replica scoped to your account, queried directly with SQL, for teams that prefer database-native access patterns.

USE CASES
How Teams Use Feed
Three systems the schema powers
Cost-aware routing
Pick the cheapest model that clears the bar
Route inference requests to the most cost-efficient SKU meeting your capability constraints. Filter on context window, tool calling and modality, then sort by normalized price.
Cost modelling
Build forecasts and projections
Ingest the time series into proprietary models and track price evolution across vendors. Build signals from cost compression patterns and competitive pricing dynamics.
Pricing embeds
Surface live rates inside your own product
Display inference pricing in your own dashboards or SDKs. License Feed as a data layer, attribute Attic Standard as the source, and keep your build focused on your differentiation.

GET STARTED
Two Paths to Live Data
Evaluate with a pilot, scale with an annual licence
01
Schedule a call
A 30-minute walkthrough of the schema, the delivery options and your specific use case. We confirm fit before any commercial conversation.
02
Pilot or annual
Start with a three-month pilot to validate integration and data fit, or move directly to an annual licence for production deployment.
03
Production integration
Wire Feed into your stack with engineering support, SLA-backed delivery, an audit trail and ongoing schema stability.
Evaluation
$25,000/ 3-month pilot
Validate integration and data fit before an annual commitment.
Production
From $100,000/ year
Full annual licence, priced on use case, delivery pattern and volume.
Three-month pilot agreement·Custom annual pricing on request·Volume discounts available

Security and compliance. Attic Standard runs on SOC 2 Type II certified platforms and ISO 27001:2022 certified storage infrastructure, and inherits those controls through its platform architecture. Our providers also maintain ISO 27017, ISO 27701 and GDPR compliance.


WHERE TO GO NEXT
Other Ways into Attic Standard
The same benchmark, delivered differently

Feed is the benchmark as raw records. MCP puts the same numbers inside the tools you already code in, and Terminal gives the full historical record on a dashboard built for analysis.


Talk to the team

Whether you need a demo, a quote, or a conversation about coverage, we respond within one business day.

ALL FIELDS REQUIRED
Enter your first name
Enter your last name
Enter a valid work email
Enter your company
MCP
Terminal
Feed
Coverage question
Methodology question
Vendor onboarding
Press
Partnerships
Other
Select an option
Enter a message
We respond within one business day.

EMAIL
team@a7om.com
General enquiries and support
ATHENS
Monday to Friday, 09:00–18:00 EET
Research and operations
NEW YORK
Monday to Friday, 09:00–18:00 ET
Commercial and press

ENTITY
Member One EOOD
RESPONSE
One business day
DEMOS
30 minutes, by appointment
PRESS
Figures may be cited with attribution
Price Indexes
Loading price indexes...