Thinking Machines has launched Inkling Small, an open-source AI model that delivers performance close to its larger predecessor while requiring only about one-quarter of the computational resources. Released on July 31, 2026, the model targets enterprises and developers who need powerful AI capabilities without the infrastructure overhead of larger frontier models.
Compact Yet Capable
Inkling Small represents a shift in the AI landscape toward efficient, deployable models. While the race for ever-larger models continues, there’s growing recognition that many real-world applications don’t require the full capabilities of frontier models—what matters is getting the right balance of performance, size, and cost.
The model achieves performance levels that are “nearing” its predecessor while being approximately 75% smaller. For companies that want control over their data, model behavior, and fine-tuning capabilities, this smaller footprint may be more important than chasing the highest possible benchmark score.
Open Source for Enterprise
The release maintains Thinking Machines’ commitment to open-source AI. Inkling Small is available under an open license, allowing enterprises to:
- Deploy on-premises or in private clouds
- Fine-tune on proprietary datasets
- Maintain full data privacy and security control
- Customize model behavior for specific use cases
The Efficiency Frontier
The release comes amid growing industry focus on AI efficiency. As model competition shifts toward cost optimization, companies are increasingly looking for models that deliver sufficient performance at a fraction of the compute cost.
Inkling Small joins a growing ecosystem of efficient open-source models that challenge the assumption that bigger is always better. For edge deployment, on-device inference, and privacy-sensitive applications, compact models like Inkling Small offer a compelling alternative to frontier models.
This release signals that the AI industry is maturing beyond pure benchmark-chasing toward practical, deployable solutions that balance capability with efficiency.