Tested, not claimed

Marketing pages assert; this page proves. Every number below is either computed live from the model registry at build time, or tied to a dated measurement described in the code. When a number ages, the page ages with it — nothing here is hardcoded marketing copy.

Live counters

What the catalog actually contains

These counters are computed from the model registry when the page is built — change the registry, the numbers change. No manual sync, no stale figure.

live chat models

served by EU sovereign providers

11/25

are Frontière Verified (6/6)

benchmark-tracked model families

Verification detail

Verification status of every live model

A 6/6 score means all six checks passed; a partial score shows how many have passed, and an unticked check was either not run yet or run without meeting the published rule; “not yet verified” means no check is on record yet. The effective provider is the route the gateway tries first.

Model	Effective provider	Checklist
qwen3.5-397b	scaleway	6/6
glm-5.2	scaleway	6/6
deepseek-v4-flash-0731	scaleway	6/6
qwen3.6-35b	scaleway	6/6
qwen3-coder-30b	scaleway	6/6
gpt-oss-120b	scaleway	6/6
qwen3.8-27b	aki	6/6
glm-5.3	aki	6/6
deepseek-v4.1-flash	aki	6/6
gemma-4-26b	aki	6/6
kolibri-1	aki	6/6
qwen3-235b	scaleway	5/6
llama-3.3-70b	scaleway	5/6
pixtral-12b	scaleway	5/6
mistral-small-3.2-24b	scaleway	5/6
apertus-v1.5-70b	infomaniak	5/6
deepseek-v4-pro	openrouter	5/6
mistral-small-4-119b	mistral	5/6
mistral-large-4	mistral	4/6
gemma-3-27b	openrouter	4/6
kimi-k3	modal	3/6
muse-spark-1.1	openrouter	3/6
muse-spark-1.2	openrouter	3/6
qwen3.8-flash	openrouter	3/6
glm-5.3-flash	modal	not yet verified

Dated evidence

Each claim, its evidence, its date

Function calling is measured, model by model

2026-10-20

Measured, never inferred: one real API call per model with a get_weather tool (plus, on part of the first run, a full role:"tool" round-trip), repeated when a model changes provider. 15 of the 25 live models have a dated measurement, from 2026-08-06 to 2026-10-20; 15 of them returned a valid tool_call with well-formed JSON arguments. The other 10 live models do not have a dated measurement on record yet.

Frontière AI stores no prompt or completion

2026-09-21

By design, not policy-only: prompts and completions are never written to our database — not from the API, not from MCP, not from the web console, whose conversations live only in your browser. We keep the token counts billing requires. On the provider side, we record what each provider documents: OVHcloud states zero data retention, Scaleway states prompts are neither stored nor logged, except requests that cause errors or abuse, kept up to two weeks. On fast-access models (Modal, OpenRouter), retention follows the provider's own policy, which we do not verify.

Sovereign models run on EU-controlled providers

2026-09-22

The 18 live models labeled EU sovereign are served by Scaleway (10), aki (5), Mistral AI (2) and Infomaniak (1) — companies with no non-EU capital control. No live model runs on Frontière-operated servers yet. Without an explicit opt-in, a sovereign model never leaves sovereign providers, fallback included. Every request still transits our application host, Vercel (US company, Paris region), which stores no prompt or response. Provider facts are transcribed from audits recorded in the registry; the effective provider per model is its priority-1 route.

Sovereignty methodology →

Published prices = provider prices × one fixed markup, VAT included

2026-08-13

Provider prices re-confirmed by direct checks (Scaleway pricing page, OVHcloud /v1/models endpoint, OpenRouter /v1/models) and converted at the documented ECB-referenced rates. maxCompletionTokens limits measured by real oversized requests per provider.

Benchmark scores are real, public figures

2026-08-26

MMLU, GPQA Diamond, LiveCodeBench and Arena Elo scores are the best publicly reported figures per model, sourced from Hugging Face model cards, official technical reports and LMSYS — not our own run, and labeled as such.

Benchmark data file →

The score

Sovereignty Score — deterministic, recomputable

The Sovereignty Score shown on every model page is a deterministic 0–100 aggregation of five weighted factors computed from registry facts. You can recompute any score yourself from the published formula — no black box.

Read the full methodology →

The source

One registry, every fact

Everything on this site — prices, jurisdictions, verification checklists, tool-calling results, context windows — comes from a single typed model registry whose comments document each verification: what was tested, when, how. It is the opposite of a marketing layer: the code is the source of truth.