Skip to main content
The Sovrnti Risk Model places every product in a market on three axes: A Snapshot is the score for every product in a market, plotted on those axes, plus the dimension bands (e.g. Foundation, Frontier) that give the numbers narrative meaning. Snapshots are the unit of analysis — you don’t score products in isolation, you score them against the field.

​
Why three axes, not one

A composite “sovereignty score” would collapse the trade-off. The whole point is that Innovation and Control pull in opposite directions for most markets today — the most innovative model is usually not the most controllable one, and vice versa. Scale exists because a frontier model with no SLA is not an option for most enterprises. Reading a Snapshot is reading a position in this three-dimensional space. There’s no single right answer; there’s a context-dependent best answer.

​
Score range

Every axis is scored 0..100. The bands are the natural language layer on top:
  • Foundation (lower band) — table stakes. The product does the job, but with limits.
  • Frontier (upper band) — defines the state of the art in the market for this axis.
Specific band cutoffs vary by market and are returned with each Snapshot under dimensions[].bands[]. Trust the band names over the raw numbers when communicating outside the team — "Le Chat is on the Control frontier" lands better than "Le Chat: control=81".

​
What feeds the score

Each axis is a weighted aggregate of market-specific signals. The full methodology is returned on every Snapshot under methodology — Innovation scored across model capability benchmarks, multimodal support, and API feature depth. Scale scored across SLA tiers and customer footprint. Control scored across EU hosting, certifications, and data residency. Methodology text is per-market. Quote it verbatim in any analyst-facing brief — paraphrasing creates drift.

​
Where to read it from

​
From the REST API

The response is a MarketSnapshot resource — products, dimensions, methodology — in a single payload. Scope: Market.Read.All.

​
From an MCP client

The MCP tool returns the same payload plus an embedded ui://sovrnti/snapshot/{marketId} HTML resource for clients that render UI artifacts (Cursor, future Sovrnti-aware clients). Scope: Market.Read.All. Cost: 5 credits. See sovrnti_market_get_snapshot for the full input/output schema.

​
Discovering valid markets

marketId is a kebab-case slug — copilots, cloud-compute, vector-databases. To list every valid value, hit GET /v1/markets (REST) or call sovrnti_markets_list (MCP). Same payload either way.

​
What the model deliberately does not do

  • No single sovereignty score. Innovation, Control, and Scale are reported separately so you can weight them yourself.
  • No vendor recommendations in the data layer. The Snapshot positions products — it doesn’t tell you which one to buy. Recommendation is the Sherpa layer’s job; it isn’t yet generally available.
  • No retroactive rescoring without a version bump. When the methodology changes materially, the next Snapshot release carries a new lastUpdated and is noted in the changelog.

​
See also