Tinker - Thinking Machines Lab
Tinker is a training API for developers
Control every aspect of model training and fine-tuning while we handle the infrastructure.
Your ideas in four functions
forward_backward
Perform a forward pass and a backward pass, accumulating the gradient.
optim_step
Update weights based on the accumulated gradient.
sample
Generate tokens for interaction, evaluation, or RL actions.
save_state
Save training progress for resumption.
Supported models
THINKING MACHINES](/content/inkling/index.html)
Inkling NEW
MoE
Inkling-SmallCOMING SOON
MoE
DEEPSEEK](https://huggingface.co/deepseek-ai)
MoE
MOONSHOT](https://huggingface.co/moonshotai)
MoE
NVIDIA](https://huggingface.co/nvidia)
MoE
Nemotron-3-Super-120B-A12B-BF16
MoE
Nemotron-3-Ultra-550B-A55B-BF16
MoE
GPT-OSS](https://huggingface.co/collections/openai/gpt-oss)
MoE
MoE
QWEN](https://huggingface.co/Qwen)
MoE
Dense
Dense
Dense
Dense
MoE
MoE
Dense
Tinker uses LoRA
LoRA fine-tunes models by training a small add-on instead of changing all the original weights.
Tinker cookbook
Explore realistic examples and common abstractions for fine-tuning language models with the Tinker API.
Tinker lets researchers focus on datasets, algorithms, and environments without the complexities of compute and infrastructure.
Tyler Griggs
Tinker lets us focus on the research, rather than spending time on engineering overhead. That's something no amount of raw GPU credits can substitute.
Ziran Yang, Yong Lin, Chi Jin
The training infrastructure has been abstracted away, which makes focusing on our data and evals far easier. Tinker has made it easy to jump into RL work.
Jason Liu
Tinker has been reliable for quickly iterating without worrying about hardware or infrastructure.
Eric Gan
FAQs
How do I access Tinker?
Join here. If you're a university or organization looking for wide scale access, contact tinker@thinkingmachines.ai.
What is Tinker and who is it for?
Tinker is a flexible API for efficiently fine-tuning open source models with LoRA. It's designed for researchers and developers who want flexibility and full control of their data and algorithms without worrying about infrastructure management.
What is LoRA and why does Tinker use it?
LoRA is an efficient approach to fine-tuning that trains a streamlined adapter instead of updating all base model weights. Our research demonstrates that with the right setup, LoRA matches the learning performance of full fine-tuning while providing more flexibility and requiring less compute.
Do I need to manage training infrastructure?
Tinker handles scheduling, tuning, resource management, and infrastructure reliability so you can focus on the training data and algorithms. Behind the scenes, Tinker orchestrates distributed training on powerful GPU clusters for efficient utilization.
What do I need to start training?
A dataset of supervised learning examples or reinforcement learning environments. After picking a base model to train on, the Tinker API provides simple functions to compute gradients, update the weights, and sample outputs from the trained model. See our cookbook for examples to get started.
What models can I train with Tinker?
Tinker is currently available for a broad selection of open-source models, ranging from compact models like Qwen3.5-4B to large MoEs like Kimi-K2.6. We plan to expand our model lineup with even more choices soon.
How do you handle my training data?
Your data is used solely to fine-tune your models. We do not use your data to train our own models.
Can I download my model weights?
Yes, we have an API endpoint that lets you download any checkpoint you've saved.
How do I stay up to date with Tinker news?
Follow Tinker on X for API updates, new model and feature announcements, and weekly roundups of featured projects from the Tinker community.
How much do I have to pay to use Tinker?
Tinker uses a pricing plan that reflects usage. All prices are in USD per million tokens. See pricing in our docs. Checkpoint storage is charged at $0.10/GB-month.