Tinker
Control every aspect of model training and fine-tuning
About
Tinker is a training API developed by Thinking Machines that allows developers and researchers to control every aspect of model training and fine-tuning, while the platform handles the underlying infrastructure. It supports a broad selection of open-source models (Qwen, Llama, GPT-OSS, DeepSeek, Moonshot) and utilizes LoRA for efficient fine-tuning. Key functionalities include `forward_backward`, `optim_step`, `sample`, and `save_state`. The service is priced per million tokens for prefill, sample, and train operations.
Categories & Tags
Color Palette
White
#FFFFFF
Dark Grey
#333333
Primary Blue
#007BFF
Typography
Sans-serif
Heading, Body
Design Review
Similar Products
Clear for Slack
Clear messages get answered quicker
Griply 2026
Achieve your goals with a goal-oriented task manager
HappyMail
We made email simple again
Blober.io
The easiest way to transfer files between cloud providers.
Supaguard
Scan, Detect & Protect Your Supabase Data
Timelines Time Tracking 4
Track your time to achieve your New Year resolutions.
SoftReveal — Reveal less. Engage more.
Hide Content, Reveal on Click
Reword
Rewrite messages without leaving your workflow
Radial
Your shortcuts, one gesture away
MoovAI
Launch viral AI ads & pro social content in minutes
Resell AI
Reselling workflow with market-based price suggestions
Qwen-Image-2512
SOTA open-source T2I model with even greater realism