DFT v1
Our flagship writing model, post-trained with Distribution Fine-Tuning to produce clear, varied prose without the repetitive patterns common to general-purpose models.
- $2.5 / M input
- $12 / M output
Product
Models
Post-training services
Train a custom model in your style
We provide custom solutions for better writing, editing, and publishing, including models trained on your style.
Distribution Fine-Tuning
We trained the Deft models with Distribution Fine Tuning (DFT), a custom training algorithm. We can do the same for you, using your data, so your LLM produces texts with a token distribution closer to your ideal.
AI Advising
We advise your team on AI transformation, including how to setup or improve AI-assisted writing and editing systems, as well as other standard model training techniques, RL environments, AI strategy, etc.
Platform
Console
Generate and rewrite in the browser, no setup required.
Open Console →API
A single /v1/generate endpoint for pipelines and products.
Read the docs →Custom Agents
The SEO Agent and Rewriter Agent, free hosted workflows.
Try an agent →