01What it sells and who buys it
Thinking Machines provides Tinker for adapting models to customer data and tasks, alongside open Inkling models. Its buyers include researchers and development teams customizing AI.
02How the business makes money
Tinker charges per million input, output and training tokens, plus monthly checkpoint storage per GB. Downloading open models and paying for managed training are distinct usage paths.
03Marketing and customer acquisition
After financing the research team, Thinking Machines linked open models to a managed training product.
- 01
Target developers customizing models
SourceTinker provides a training environment for task-specific models. Inkling offers evaluation through both downloadable models and managed experimentation.
- 02
Financing the next stage
SourceA $2 billion seed round led by a16z in July 2025 valued the company at $12 billion. Initial financing established research and computing capacity, followed by the release of developer training products.
- 03
How distribution and adoption expanded
SourceOpen models are linked to Hugging Face downloads and direct customization through Tinker. Model cards, benchmarks and training documentation help developers evaluate the technology and run a first experiment. Managed training provides a commercial route from that research visibility.
How the growth mechanism works
- Read the research
- Test the open model
- Customize through Tinker
- Expand compute usage
Open models build technical credibility; managed training reduces execution work. Early financing backed the team and research capability and must be read separately from later paid adoption.
04Funding and expansion
A $2 billion seed round led by a16z in July 2025 valued the company at $12 billion. Initial financing established research and computing capacity, followed by the release of developer training products.