Building blocks · by Thinking Machines Lab

Tinker

Custom-train an open AI chatbot with your own examples or rewards.

data, text → deployed, data · text → text
Way in
Official APISign up yourself, api key
To start
$none stated
Runs
In the cloud
Checked
Nov 4, 2026Prices as of that day

What it can do 2

Fine-tune or RL-train an open-weight LLM with your own loop (SFT, DPO, RL with custom rewards)~25 base models; some retire on short notice (three retired 2026-10-23)data, text → deployed, dataper 1M tokens, e.g. Qwen3-8B $0.44 train / $0.60 sample; Kimi-K2.6 $4.84 train (32K ctx) or $15.40 (128K) / $5.49 sample (pricing page, as of 2026-10-19)
Sample from trained checkpoints; serverless inference (beta, Inkling models only)serverless inference beta covers Inkling and Inkling-Small onlytext → textper 1M tokens; checkpoint storage $0.10/GB/mo

Does the same job

Prices and limits as of Nov 4, 2026, from Thinking Machines Lab’s own pages. They change; check before you rely on them.Not affiliated with Thinking Machines Lab. Not advice.Something wrong?
Made by Jasper Burns

A short note when I ship something new, about once a month. One list for all my projects.