中文
AI tools

Coding & apps

Thinking Machines Lab

Thinking Machines Lab is an AI research and product company. It offers the open-weight Inkling model family and Tinker, a platform for fine-tuning and training models.

Visit website

Offers

Inkling 2 Small is free on OpenRouter and OpenCode Zen

Inkling 2 Small is free for a limited time on OpenRouter and OpenCode Zen, with about 1 million tokens of context

Eligibility
An OpenRouter account can call the :free route; the OpenCode free tier has to be used inside OpenCode. The OpenCode route is not available in the EU, Iceland, Liechtenstein, Norway, Cuba, Iran, North Korea, Russia, Ukraine, or Venezuela. The free routes may retain request content to improve the model, so do not upload confidential or personal data.
Benefit duration
Free for a limited time; no end date has been announced
Claim offerSourceRead details

What you can do

Researchers and developers who want to try or fine-tune Inkling models, or train custom models on Tinker.

Get started

Open https://thinkingmachines.ai/ to learn about the company and its products. To use the Tinker training platform, start at https://thinkingmachines.ai/tinker/, and create an account at https://auth.thinkingmachines.ai/sign-up if you need one. After you sign in, you can try models in the Tinker Playground at https://tinker.thinkingmachines.ai/playground. The getting-started docs are at https://tinker-docs.thinkingmachines.ai/tinker/quickstart/. Inkling is their model family. The public models are currently Inkling, Inkling-Small, and Inkling 2 Small. Inkling 2 Small is the first model in the second generation. It is a mixture-of-experts model with about 594 billion total parameters and 17 billion active parameters, accepts text and image inputs, and supports adjustable reasoning effort. The official model card states a context window of up to 1 million tokens, and the weights are released under the Apache 2.0 license. An update on November 12, 2026 said the weights are available to download on Hugging Face, and that the model can also be called via API on NVIDIA, Fireworks, Together, Baseten, Nebius, Modal, and Cloudflare. You can also fine-tune and sample it on Tinker.

Sources