Comparison
Frontière AI vs LowRouter: two French LLM routers
LowRouter and Frontière AI are both French, both OpenAI-compatible and both prepaid — and on the same capacity, LowRouter is cheaper. Carbonifer SAS, a Nantes company trading as LowRouter, bills tokens at the upstream provider's own price with no per-token markup and charges a fee on each top-up instead: on Qwen3 235B served by Scaleway in Paris, that is €0.75 per million input tokens before VAT and fee, against our €1.05 before VAT (€1.26 VAT included) for the same model on the same provider. The real difference is the default. LowRouter routes to EU and US providers alike — including OpenAI, Anthropic, AWS Bedrock and DeepInfra, which its own privacy policy says remain subject to US law, including the CLOUD Act — and a request without a region preference goes wherever its router picks on availability and carbon intensity, unless you pin a region in the model ID. Frontière AI is a French company: its 17 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 you want the widest catalog at upstream prices and will pin regions yourself, LowRouter fits; if you want the jurisdiction decided per model before you call, that is what we are built for.
Side by side
Every row below comes from the other side's own published pages, checked on 2026-10-12 and linked at the end of this section — not from a third-party comparison.
| Criterion | Frontière AI | LowRouter |
|---|---|---|
| Operating company | FLEECE AI SASU | Carbonifer SAS (trading as LowRouter) |
| Headquarters | France | Nantes, France |
| Governing law | French / EU law | French law (its terms, section 15) |
| US CLOUD Act exposure | Inference on the 17 sovereign models runs at EU companies; requests transit the gateway on Vercel (US company, Paris region, no prompt or response stored) | The company is French; its privacy policy says its US-established providers remain subject to US law, including the CLOUD Act |
| How the catalog works | Curated catalog, 24 models live, each with a named primary provider | Router over EU and US providers, closed frontier models included; upstream price per token plus a fee on each top-up |
| Qwen3 235B, per 1M tokens (in / out) | €1.26 / €3.78 incl. VAT (€1.05 / €3.15 excl. VAT) | €0.75 / €2.25 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 | Not retained after the request completes, except where strictly necessary for security, abuse prevention or legal compliance (privacy policy); call metadata kept for usage reports |
| Jurisdiction visible per model | Yes — sovereign / fast access, on the card and in the API response | Each route names provider, model and region; a region pinned in the model ID fails closed, otherwise its router picks on availability and carbon intensity |
Both sides publish in euros, so nothing is converted. LowRouter's figure is the Scaleway fr-par route in its models API — the same provider as our primary route for this model. LowRouter's pricing page says tokens are billed at the upstream provider's own price with no per-token markup, with VAT added at checkout (reverse charge for businesses with a verified EU VAT number) and a fee of 5.5 % to 8.5 % plus €0.30 on each top-up (checked on 2026-10-12). Frontière AI publishes VAT included, with no top-up fee — so the price verdicts on this page compare both sides excluding VAT and before LowRouter's fee (our price ÷ 1.2, French VAT).
Why the price gap? Same model, same provider: LowRouter passes Scaleway's own price through and earns its fee on top-ups, while Frontière AI serves this model through Scaleway, an EU sovereign provider with a fixed margin in the per-token price and no top-up fee. Excluding VAT on both sides (Frontière AI: €1.05 in / €3.15 out per million), on this model LowRouter is cheaper on both input and output. Even with LowRouter's top-up fee added, it remains cheaper on this model.
Sources, checked on 2026-10-12: api.lowrouter.ai — models & prices (Qwen3 235B, Scaleway fr-par) · lowrouter.ai — pricing · lowrouter.ai — privacy · lowrouter.ai — sub-processors · lowrouter.ai — sovereignty & transparency · lowrouter.ai — terms · lowrouter.ai — legal notice
Where LowRouter is the better choice
- Cheaper on the same capacity: upstream price per token, no per-token markup, and a fee of 5.5 % to 8.5 % plus €0.30 per top-up instead.
- A far wider catalog, including closed frontier models from OpenAI, Anthropic and Google, and routes on the EU regions of US hyperscalers.
- Region pinning per request in the model ID, failing closed, with the chosen region reported and carbon figures per route.
- Reverse charge for EU businesses with a verified VAT number. Frontière AI's prices include French VAT, and it does not apply reverse charge yet.
Where Frontière AI is the better choice
- The default is decided per model, not per request: a model labeled EU sovereign stays with sovereign providers unless you explicitly allow others — nothing to pin, nothing to forget.
- 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.
- No top-up fee: the amount you pay is the amount credited, and the per-token price you read is the one you are billed.
- A curated catalog with tool calling measured model by model, rather than more than a thousand provider routes to sort through.
How to choose, concretely
Both are OpenAI-compatible, so switching is a base_url and a key:
- Lowest price on a given provider, closed frontier models, or carbon-aware routing: LowRouter.
- Regulated workloads where nobody should have to remember to pin a region: Frontière AI's sovereign-labeled models.
- Either way, check what each route actually is before you send personal data: LowRouter's sub-processor list and our model labels both say it.
FAQ
Is LowRouter cheaper than Frontière AI?
On the model we compared, yes. Qwen3 235B on Scaleway costs €0.75 in / €2.25 out per million tokens on LowRouter before VAT and before its top-up fee, against €1.26 / €3.78 VAT included on Frontière AI. Excluding VAT on both sides (Frontière AI: €1.05 in / €3.15 out per million), on this model LowRouter is cheaper on both input and output. LowRouter adds 5.5 % to 8.5 % plus €0.30 per top-up, which does not close the gap.
Both companies are French — are both outside the reach of US law?
The companies are not directly subject to the CLOUD Act, but the providers they route to are what matters. LowRouter's privacy policy says its US-established providers remain subject to US law, including the CLOUD Act. The same logic applies on our side: Frontière AI's fast-access models are served by US companies and labeled as such; only the models labeled EU sovereign run at providers outside its direct reach.
How do I keep LowRouter requests in the EU?
By pinning a region in the model ID, which its sovereignty page says fails closed if the model is not served there. Without a preference, its router picks on availability and carbon intensity. On Frontière AI you pick a model labeled EU sovereign, and it stays with sovereign providers unless you explicitly allow others.
Do LowRouter and Frontière AI store prompts?
Neither says it does. LowRouter's privacy policy says it does not retain prompts or completions after the request completes, except where strictly necessary for security, abuse prevention or legal compliance. Frontière AI never writes API prompts or completions to its database — only the token counts billing requires.
See what we actually serve
24 models live, each with its publisher, licence, hosting jurisdiction and real price.