Brie

Model Family · November 2025

Small language models that think and write in the register of continental philosophy, critical theory, and speculative fiction. Not retrieval — engagement.

Models

Brie · Qwen 2.5 3B

Highest specialization

91.2%

in-domain win rate

Deepest philosophical alignment. Aggressive domain specialization while retaining general competence. Recommended starting point.

huggingface.co →

Brie · Llama 3.2 3B

Best general preservation

80.4%

in-domain win rate

Strong domain performance without sacrificing breadth. Best for use cases where general capability still matters.

huggingface.co →

Brie · Qwen 2.5 0.5B

Consumer hardware

72%

comprehensive win rate

Runs locally on an M4 MacBook at negligible cost. Proof that domain adaptation works at sub-1B scale.

huggingface.co →

Intellectual territory

Continental philosophy

Phenomenology

Existentialism

Critical theory

Speculative reasoning

Literary analysis

Cultural criticism

Creative essays

How it was made

Trained on 1,213 examples authored through iterative discussions — not scraped, not synthetically generated. The dataset took years to develop; the training cost $3. Full methodology →

Links