Tiny
50M – 125MPhones, browsers, embedded devices
Small enough to run inside a web page or on low-power devices. Fast and inexpensive, and strong on narrow tasks once fine-tuned.
Français English
Aulvero trains compact bilingual models for French and English, from fifty million to nine hundred million parameters. They're built to run on the hardware people already own.
Most people start with the hardware they have, not a parameter count. So that's how we group the models.
Phones, browsers, embedded devices
Small enough to run inside a web page or on low-power devices. Fast and inexpensive, and strong on narrow tasks once fine-tuned.
Phones and laptops
The everyday middle. Runs comfortably on a phone or any laptop CPU, with enough capacity for general instruction following.
Laptops and single GPUs
Our most capable models. Still small enough for one consumer GPU or a recent laptop, with noticeably stronger reasoning and writing.
Each model ships as a Base checkpoint and an Instruct checkpoint fine-tuned from it. A Code variant comes later.
The raw pre-trained model. It continues text rather than following instructions. Use it as the starting point for your own fine-tuning.
Fine-tuned from Base to follow instructions and hold a conversation in French or English. Most people should start here.
Base models further pre-trained on a code-heavy mix with fill-in-the-middle, for completion and repository understanding.
One architecture and one codebase cover every size. The same four steps take each model from raw text to a download.
French web text from FineWeb-2, English educational text from FineWeb-Edu, and permissively licensed source code. Everything is cleaned, filtered and de-duplicated.
One 49,152-token tokenizer for both languages and codeEach size trains from scratch on roughly a hundred tokens per parameter. A warmup-stable-decay schedule lets us extend promising runs without starting over.
Produces the Base checkpointsBase models are fine-tuned on French and English conversations, and learn only from the assistant's side of each exchange.
Produces the Instruct checkpointsWeights are exported in a standard Hugging Face format, so they load directly in Transformers, vLLM, llama.cpp and Ollama.
No custom code needed to run a model# one command per stage, the same for every size
$ make prepare
$ make pretrain MODEL=kestrel-250m GPUS=8
$ make finetune MODEL=kestrel-250m GPUS=8
$ make export MODEL=kestrel-250mThe names follow birds, smallest to largest. The rest of the name says exactly what you're downloading.