Inkling Small AI Model Debuts
Thinking Machines' new open-source AI model achieves near-parity with its predecessor at a quarter of the size
Thinking Machines has introduced Inkling Small, an open-source AI language model that closely matches the performance of its larger predecessor, Inkling, despite being significantly smaller. Inkling Small has 276 billion parameters, compared to Inkling's 975 billion parameters. The new model supports text, image, and audio inputs, produces text outputs, and has a context window of up to one million tokens.
The smaller size of Inkling Small reduces compute requirements, inference costs, and deployment footprint, making it an attractive option for enterprises. Despite its smaller size, Inkling Small preserves much of the coding, reasoning, and multimodal performance of the original Inkling model.
Inkling Small uses 12 billion active parameters per token, compared to Inkling's 41 billion active parameters. The model has been released under a permissive Apache 2.0 license, allowing developers to use and modify it freely.
Why it matters
The release of Inkling Small demonstrates the ongoing progress in developing more efficient and effective AI models. As AI models continue to grow in size and complexity, reducing their computational requirements while maintaining performance is crucial for widespread adoption. This development has significant implications for the field of artificial intelligence, particularly in areas such as natural language processing and multimodal reasoning.
The release of Inkling Small demonstrates the ongoing progress in developing more efficient and effective AI models.
What you can learn from this
- Model pruning and efficiency: Inkling Small's ability to match its predecessor's performance at a quarter of the size demonstrates the importance of model pruning and efficiency techniques in AI development. Learners can explore how model pruning works, including techniques such as reducing parameter counts and using knowledge distillation to transfer knowledge from larger models to smaller ones.
- Multimodal reasoning: Inkling Small's support for text, image, and audio inputs highlights the growing importance of multimodal reasoning in AI. Learners can delve into the concepts and techniques behind multimodal reasoning, including how to integrate multiple input modalities and generate coherent outputs.
- Open-source AI development: The release of Inkling Small under an open-source license showcases the value of collaborative development in AI. Learners can learn about the benefits and challenges of open-source AI development, including how to contribute to and modify existing models, and how to navigate licensing and intellectual property issues.
- Compute requirements and deployment: The reduced compute requirements and deployment footprint of Inkling Small underscore the need for AI developers to consider the practical implications of their models. Learners can explore strategies for optimizing AI models for deployment, including model compression, quantization, and knowledge distillation.
- Benchmarking and evaluation: The use of third-party benchmarks, such as the Artificial Analysis Intelligence Index, to evaluate Inkling Small's performance demonstrates the importance of rigorous testing and evaluation in AI development. Learners can learn about different benchmarking methodologies and how to design and implement effective evaluation protocols for AI models.
We teach this
Sources
- Thinking Machines debuts Inkling Small open source AI model nearing performance of predecessor at about 1/4 size — VentureBeat
Our reporting is an original summary; full coverage is at the links above.
Don't just read about it — build it.
Square 1 teaches the skills behind the headlines, with every line of your work graded by AI. Find your starting point in 3 minutes.
Get your free skill report