Inkling Small
The smaller, cheaper-to-serve member of the Inkling family.
Inkling Small is the efficient model in Thinking Machines Lab’s open Inkling family: a multimodal mixture of experts with 12B of 276B parameters active.
- Provider
- Thinking Machines Lab
- Type
- Language model
- Released
- Jul 30, 2026
- Input
- text, image, audio
- Output
- text
- Open weights
- Yes
- License
- Apache 2.0
Best for
- Reasoning
- Coding
- Agents
- High volume / low cost
Strengths
- Apache 2.0 weights
- 12B active parameters
More from Thinking Machines Lab
- Inkling — Thinking Machines Lab. Thinking Machines’ first open-weight frontier MoE.