Tinker

Train LLMs with
the Tinker API

A cloud training API that gives you full control over your data, algorithms, and models. Write training loops locally — Tinker handles the GPUs.

uv pip install tinker tinker-cookbook
Training

LoRA Fine-Tuning

Model-agnostic LoRA training from 1B to 1T+ parameters — dense and MoE architectures, text and vision.

Algorithms

Full Control

SFT, RL (GRPO, PPO), DPO, distillation. Write your own loop or use cookbook abstractions.

Infrastructure

Remote GPU Clusters

Distributed training via API. Author code on a CPU machine — Tinker runs the compute.

Examples

15+ Recipes

End-to-end recipes for math RL, code RL, tool use, multi-agent training, and more.

28+
Supported Models
15+
Training Recipes
23
Interactive Tutorials