Evidence-backed European AI register
FlexAI
Reviewed identity, roles, model listings, pricing and evidence for FlexAI.
Dataset compiled 2026-08-27 · evidence over inference
Reviewed identity
- Headquarters scope
- EU headquarters
- Headquarters record
- Paris
- Legal entity
- SAS (exact entity not verified)
- Ownership
- VC-backed private
- Founded
- 2023
- Confidence
- high
- Last verified
- 2026-08-22
Evidence boundaries
Organisation listings identify a reviewed canonical record. Model listings show an association in the source register; they do not by themselves establish model creation, current API availability, hosting location or infrastructure ownership.
Model listings
- Type
- text
- Parameters
- 12B
- Context
- unknown
- Languages
- multilingual
- Licence
- Apache-2.0
- Release
- 2024-07
- Type
- text
- Parameters
- 8B
- Context
- unknown
- Languages
- multilingual
- Licence
- Llama 3.1 Community License
- Release
- 2024-07
- Type
- text
- Parameters
- 70B
- Context
- unknown
- Languages
- multilingual
- Licence
- Llama 3.3 Community License
- Release
- 2024-12
- Type
- text
- Parameters
- 20B MoE
- Context
- unknown
- Languages
- multilingual
- Licence
- Apache-2.0
- Release
- unknown
- Type
- text
- Parameters
- 120B MoE
- Context
- unknown
- Languages
- multilingual
- Licence
- Apache-2.0
- Release
- unknown
- Type
- multimodal
- Parameters
- 26B MoE (4B active)
- Context
- unknown
- Languages
- multilingual
- Licence
- Gemma licence
- Release
- unknown
- Type
- multimodal
- Parameters
- 31B
- Context
- unknown
- Languages
- multilingual
- Licence
- Gemma licence
- Release
- unknown
- Type
- code
- Parameters
- 30B MoE (3B active)
- Context
- unknown
- Languages
- code
- Licence
- Apache-2.0
- Release
- unknown
- Type
- multimodal
- Parameters
- 35B MoE (3B active)
- Context
- unknown
- Languages
- multilingual
- Licence
- Apache-2.0 (Qwen)
- Release
- unknown
- Type
- multimodal
- Parameters
- 9B
- Context
- unknown
- Languages
- multilingual
- Licence
- Apache-2.0 (Qwen)
- Release
- unknown
- approx 182 further models
- Type
- text
- Parameters
- unknown
- Context
- unknown
- Languages
- unknown
- Licence
- unknown
- Release
- unknown
Published token prices
| Model | Input €/1M | Output €/1M | Currency | Evidence |
|---|---|---|---|---|
| Gemma 4 26B A4B | €0.0513 | €0.2564 | USD | price list |
| Gemma 4 31B IT | €0.0769 | €0.2906 | USD | price list |
| GPT-OSS 120B | €0.0333 | €0.0855 | USD | price list |
| GPT-OSS 20B | €0.0256 | €0.1111 | USD | price list |
| Llama 3.1 8B Instruct | €0.0171 | €0.0256 | USD | price list |
| Llama 3.3 70B Instruct | €0.0855 | €0.2735 | USD | price list |
| Mistral Nemo | €0.0154 | €0.0256 | USD | price list |
| Qwen3 Coder 30B A3B | €0.0598 | €0.2222 | USD | price list |
| Qwen3.5 9B | €0.0855 | €0.1282 | USD | price list |
| Qwen3.6-35B-A3B | €0.0855 | €0.1282 | USD | price list |
Profile note
The closest thing France has to a European Together/Fireworks: an OpenAI-compatible 192-model per-token endpoint with a real free tier, founded by ex-Intel CTO Brijesh Tripathi and ex-NVIDIA Dali Kilani, and the only provider in this survey that cleanly spans tokens → dedicated GPU-hours → private AI factory. Its floor of USD 0.018/1M input (Mistral Nemo) is the cheapest verified per-token rate in this entire directory segment. The caveat is that it is an abstraction layer, not an infrastructure owner — datacentre locations and EU residency of the compute are not published, so it should not b…