Pricing
What Frontière AI costs: per token, prepaid, margin and VAT included
Frontière AI charges per token, with the margin and VAT already included, and no subscription. You top up a prepaid balance from €10 and every successful call draws it down; once the balance is used up, new calls are refused. The prices below are what you actually pay, shown at 100,000 tokens — the amount debited from your balance when your prompts consume that many.
All prices are per 100,000 tokens, in EUR, margin and VAT included.
Prepaid, no subscription
There is no monthly plan and no card on file. You top up once (from €10) and use the balance until it runs out — no bill to pay at the end of the month, no surprise when attention drops. Once the balance is at or below zero, new calls return HTTP 402. A call already running when the balance runs out still completes and is billed, so the balance can dip slightly below zero; that amount comes off your next top-up.
Chat & reasoning models
Every chat model is billed by input and output tokens at its own rate. Reasoning models consume output tokens (at the output rate) while thinking, before any visible text.
| Model | Served by | Price /100k in | Price /100k out |
|---|---|---|---|
| Qwen3 235B (instruct)qwen3-235b | EU sovereign · Scaleway | €0.126 | €0.378 |
| Qwen3.5 397B (multimodal)qwen3.5-397b | EU sovereign · Scaleway | €0.1008 | €0.6048 |
| GLM-5.2 (Zhipu/Z.ai)glm-5.2 | EU sovereign · Scaleway | €0.3024 | €0.924 |
| DeepSeek V4 Flash 0731deepseek-v4-flash-0731 | EU sovereign · Scaleway | €0.0672 | €0.1344 |
| Llama 3.3 70Bllama-3.3-70b | EU sovereign · Scalewayfallback: OVHcloud | €0.1512 | €0.1512 |
| Qwen3.6 35B A3B (multimodal)qwen3.6-35b | EU sovereign · Scaleway | €0.042 | €0.252 |
| Qwen3 Coder 30Bqwen3-coder-30b | EU sovereign · Scaleway | €0.0336 | €0.1344 |
| Pixtral 12B (vision)pixtral-12b | EU sovereign · Scaleway | €0.0336 | €0.0336 |
| GPT-OSS 120B (OpenAI)gpt-oss-120b | EU sovereign · Scalewayfallback: OVHcloud | €0.0252 | €0.1008 |
| Mistral Small 3.2 24Bmistral-small-3.2-24b | EU sovereign · Scalewayfallback: OVHcloud | €0.0252 | €0.0588 |
| Mistral Large 4 (preview, 512K context)mistral-large-4 | EU sovereign · Mistral AI | €0.2514 | €0.7725 |
| Apertus v1.5 70B (Swiss open model)apertus-v1.5-70b | EU sovereign · Infomaniak | €0.1294 | €0.462 |
| Kimi K3 (2.8T parameters, 1M context)kimi-k3 | Fast access (US) · Modalfallback: OpenRouter | €0.4536 | €2.268 |
| Muse Spark 1.1 (1M context)muse-spark-1.1 | Fast access (US) · OpenRouter | €0.205 | €0.6956 |
| Muse Spark 1.2 (1M context)muse-spark-1.2 | Fast access (US) · OpenRouter | €0.205 | €0.6956 |
| GLM-5.3 Flash (Zhipu/Z.ai)glm-5.3-flash | Fast access (US) · Modalfallback: OpenRouter | €0.0689 | €0.2268 |
| Qwen3.8 27B (multimodal)qwen3.8-27b | EU sovereign · AKI.IO | €0.0252 | €0.1092 |
| Qwen3.8 Flash (Qwen)qwen3.8-flash | Fast access (US) · OpenRouter | €0.0252 | €0.0773 |
| DeepSeek V4 Pro (1.6T MoE)deepseek-v4-pro | Fast access (US) · OpenRouter | €0.2453 | €0.5124 |
| GLM-5.3 (Zhipu/Z.ai)glm-5.3 | EU sovereign · AKI.IOfallback: Scaleway · US fallback on opt-in only: OpenRouter | €0.3024 | €0.924 |
| Gemma 3 27B (Google)gemma-3-27b | Fast access (US) · OpenRouter | €0.0135 | €0.074 |
| DeepSeek V4.1 Flash (552B MoE, 1M context)deepseek-v4.1-flash | EU sovereign · AKI.IOfallback: Scaleway · US fallback on opt-in only: Modal and OpenRouter | €0.1008 | €0.3024 |
| Gemma 4 26B (A4B MoE)gemma-4-26b | EU sovereign · AKI.IO | €0.0168 | €0.084 |
| Kolibri 1 (Aleph Alpha)kolibri-1 | EU sovereign · AKI.IO | €0.0252 | €0.0924 |
| Mistral Small 4 119Bmistral-small-4-119b | EU sovereign · Mistral AI | €0.0278 | €0.1109 |
Embedding models
Embeddings are billed on input tokens only — an embedding produces no output tokens, so there is no output rate.
| Model | Dimensions | Context | Price /100k in |
|---|---|---|---|
| BGE Multilingual Gemma2bge-multilingual-gemma2 | 3584 | 8,192 tokens | €0.0168no output rate |
| Qwen3 Embedding 8Bqwen3-embedding-8b | 4096 | 40,960 tokens | €0.0168no output rate |
Every figure above is the client price — the provider cost passed through Frontière AI's margin — at the 100k display unit, VAT included. This is the exact amount debited from your balance, and nothing is added at checkout.
Image models
Images are billed per second of generation, measured by the gateway and rounded up — the way our provider, Infomaniak (Switzerland), bills us. About 2 seconds for a 1024×1024 FLUX image, 5 to 6 seconds for a PhotoMaker portrait (measured 7 October 2026); each response states the seconds billed.
| Model | Served by | Sizes | Price per second |
|---|---|---|---|
| FLUX.1 [schnell] (text to image)flux-schnell | Infomaniak | 1024x1024, 1792x1024, 1024x1792 | €0.0092per second |
| PhotoMaker V2 (personalised portraits)photomaker-v2 | Infomaniak | 1024x1024, 1792x1024, 1024x1792 | €0.0092per second |
A curated selection, priced from published tariffs
Each model entered the catalog only after we checked who serves it; a detail we have not verified yet, such as a licence, is left blank on its page rather than guessed. Its cost comes from a published tariff: for most models, the price list of the provider that serves it (AKI.IO, Infomaniak, Mistral AI, Scaleway, Modal, and OpenRouter), converted once to euros when the entry is written if it is in dollars. Two cases differ: models served on a dedicated Modal endpoint, which Modal bills by the GPU-second rather than per token, are priced at the public per-token tariff of the same model on OpenRouter; and a model without a public tariff yet can carry an estimated cost. The price you pay is always that cost times one fixed markup, VAT included — the same function the billing engine uses, never a number retyped by hand on a page.
FAQ
Is there a free tier or trial period?
No. Frontière AI is prepaid only — there is no free tier and no trial period. You top up from €10 and pay per token from your balance.
How is a call billed exactly?
Token-exact, from the usage object the provider reports for your call: prompt tokens at the model's input rate, completion tokens (reasoning included) at the output rate. When a provider serves part of your prompt from its cache and reports it, those tokens are billed at the model's cheaper cache rate.
Can I top up any amount?
From €10 upward. There is no subscription and no card kept on file — you decide how much to prepay and stop whenever you want.
Are the prices VAT-inclusive?
Yes. Every price shown — top-up amounts and per-token rates alike — includes VAT where VAT applies, and nothing is added at checkout: a €10 top-up charges €10 and credits €10 of usable balance. FLEECE AI SASU's intra-EU VAT number is FR45932329188. An invoice for a top-up is issued on request; if it is for a business, send the company name and VAT number with the request.
Why do I pay more than the provider's list price?
Frontière AI routes each call to the provider that serves the model — AKI.IO, Infomaniak, Mistral AI, OVHcloud, and Scaleway for EU sovereign models, Modal and OpenRouter (US-based) for fast-access models — and bills a margin on each call; that is how the service is funded. Our prices also include VAT, while provider list prices are generally quoted excluding VAT. The price shown on every page already includes both, so there is no hidden surcharge later.
Where are the prices shown in euros per 100k tokens?
That is the display unit for every human-facing page. The machine-readable catalog (GET /api/v1/models) and the JSON-LD express the same tariff at the market's larger reference scale, so the figure reads differently but the price is unchanged.
Know what you'll pay before you build
Top up €10, change two lines in your existing client, and call any live model in minutes.