Comparison
Frontière AI vs OpenRouter: the EU-sovereign alternative
OpenRouter and Frontière AI solve different halves of the same problem, and on price OpenRouter usually wins. OpenRouter, Inc. is a New York company that routes each request to whichever provider in its pool bids lowest, across a catalog far larger than ours — on a commodity model like Llama 3.3 70B its floor price is roughly €0.087 ($0.10) per million input tokens, against our €1.26 — both before tax (€1.512 VAT included). What it cannot change is which government can compel it: a US-incorporated company answers to a US court order under the CLOUD Act, wherever the compute physically runs. Frontière AI is a French company: its 19 sovereign models run at EU providers with no non-EU capital control, the others are labeled fast access (US) before you call them, and it stores no prompts at all. If cost per token is your only criterion, use OpenRouter. If you have to answer for where your data can be compelled, that is a question about corporate control, and it is the one we are built around.
Side by side
Every row below comes from the other side's own published pages, checked on 2026-08-06 and linked at the end of this section — not from a third-party comparison.
| Criterion | Frontière AI | OpenRouter |
|---|---|---|
| Operating company | FLEECE AI SASU | OpenRouter, Inc. |
| Headquarters | France | New York, United States |
| Governing law | French / EU law | State of New York |
| US CLOUD Act exposure | Inference on the 19 sovereign models runs at EU companies; requests transit the gateway on Vercel (US company, Paris region, no prompt or response stored) | Yes — US-incorporated company, regardless of where inference runs |
| How the catalog works | Curated catalog, 26 models live, each with a named primary provider | Router across a large pool of third-party providers, price-driven |
| Llama 3.3 70B, per 1M tokens (in / out) | €1.512 / €1.512 incl. VAT (€1.26 / €1.26 excl. VAT) | €0.087 ($0.10) / €0.277 ($0.32) excl. tax |
| Prompt storage | Never stored by Frontière AI — only token counts and the amount charged. Sovereign providers: OVHcloud zero retention; Scaleway none except error or abuse requests, up to two weeks; fast-access models: the serving provider's own policy | OpenRouter's own retention is not stated in its privacy docs; downstream provider policies vary and are filterable |
| Jurisdiction visible per model | Yes — sovereign / fast access, on the card and in the API response | Per-provider data-policy table, and provider filters you set yourself |
OpenRouter's figure is the floor of a pool: the request goes to whichever provider bids lowest under the filters you set, so the effective price and the data policy both depend on who serves it that day.
Prices published in USD, shown converted at 1 USD = 0.8655 EUR (rate of 2026-08-05) alongside the original figure. OpenRouter's pricing FAQ says its prices are exclusive of applicable taxes, with VAT/GST added on invoices where required (checked on 2026-09-23). Frontière AI publishes in EUR, margin and VAT included — the amount actually debited — so the price verdicts on this page compare both sides excluding VAT (our price ÷ 1.2, French VAT).
Why the price gap? OpenRouter routes each request to whichever provider bids lowest in its pool — usually US commodity capacity at the absolute market floor. Frontière AI serves this model through Scaleway, an EU sovereign provider, with a fixed margin. The gap is not arbitrary: EU sovereign compute costs more than US commodity GPU time, and one fixed price per model over named providers costs more than an auction across dozens of discounting providers. You pay for jurisdiction certainty, not a higher rate on equivalent hardware.
Sources, checked on 2026-08-06: openrouter.ai — Llama 3.3 70B · openrouter.ai — pricing (taxes, 2026-09-23) · openrouter.ai — privacy & logging · openrouter.ai — terms
Where OpenRouter is the better choice
- A far larger catalog, including closed frontier models we do not serve at all.
- Usually cheaper per token on commodity open-weight models — sometimes by an order of magnitude, because the pool floor is set by whoever is discounting hardest that week.
- Automatic failover across a large pool of providers. Ours is narrower: 7 of our 26 models have a backup route, always to another named provider; a model labeled sovereign only falls back to another sovereign provider unless you explicitly allow others, and the x-routed-provider and x-fallback-provider response headers say which provider served each call.
- EU in-region routing (eu.openrouter.ai) is available on its Business and Enterprise plans (OpenRouter docs, checked 2026-09-24) — worth asking about if data location, rather than corporate jurisdiction, is what your policy requires. On 2026-09-24 it covered 61 of the 458 models listed, mostly closed models from OpenAI, Anthropic and Google served from the EU regions of Azure, AWS Bedrock and Google Vertex, at about 10% above their standard price.
Where Frontière AI is the better choice
- The operating company is French. There is no US parent, licensee or affiliate through which a US order could reach your data on the sovereign models.
- Every model carries its jurisdiction before you call it, in the catalog and in the API's own response — you never have to reconstruct who served a given request.
- API prompts and completions are never written to our database — only the token counts billing requires. The web console follows the same rule: conversations stay in your browser.
- The price you read is the price you are billed: one price per model, whichever of its named providers serves the call, with no per-request variance to reconcile later.
How to choose, concretely
The two are not mutually exclusive, and for most teams the honest answer is to use both:
- Regulated, personal or client data — anything a DPO signs off on: pick the models whose operator jurisdiction is settled, not just the region.
- Volume experimentation, evaluation sweeps, hobby projects, closed frontier models: the cheaper, broader pool wins and the jurisdiction question usually does not arise.
- If you only remember one rule: data-centre location is a geography question, and a subpoena is a corporate-control question. Answer the second one first.
FAQ
Is OpenRouter GDPR-compliant?
GDPR compliance and US jurisdictional reach are two different questions, and passing one does not answer the other. OpenRouter offers EU in-region routing on its Business and Enterprise plans, which addresses data location. It does not change the fact that OpenRouter, Inc. is a New York company and therefore subject to US legal process under the CLOUD Act — that is a question about who controls the company, not about where the servers are.
Is Frontière AI cheaper than OpenRouter?
Usually not, and we are not going to pretend otherwise. On Llama 3.3 70B, OpenRouter's pool floor is around €0.087 ($0.10) in / €0.277 ($0.32) out per million tokens before tax, against €1.512 / €1.512 VAT included on Frontière AI. Excluding VAT on both sides (Frontière AI: €1.26 in / €1.26 out per million), on this model OpenRouter is cheaper on both input and output. We route to a small set of named providers — EU ones for the sovereign models — and add a fixed margin; a router that auctions each request across dozens of providers will beat that on commodity models most of the time.
Can I use OpenRouter and Frontière AI together?
Yes, and it is often the right architecture. Both expose an OpenAI-compatible endpoint, so the only thing that changes between them is base_url and the key. Sending regulated workloads to the sovereign models and everything else to the cheapest pool is a reasonable split — and it is easier to justify to a DPO than one blended vendor.
Does OpenRouter let me control which provider serves my request?
Yes. OpenRouter publishes a per-provider table showing retention and training practices, lets you disable providers that train on prompts account-wide, and lets you restrict an individual request to providers meeting a given data policy. The difference is one of default: on Frontière AI each model has a named primary provider and its jurisdiction is stated up front, and a model labeled sovereign stays with sovereign providers unless you explicitly allow others (the X-Data-Sensitivity: insensitive header, or an API key set to allow all providers). Every response names the provider that served it.
See what we actually serve
26 models live, each with its publisher, licence, hosting jurisdiction and real price.