Rebellions, SqueezeBits to co-work on AI model compression

South Korea’s fabless AI chip startup Rebellions Inc. and AI model compression startup SqueezeBits announced on Wednesday that they partnered to develop technology for optimizing generative AI model compression for neural processing units (NPUs).

The two startups will focus on small language models (SLMs) compression, smaller than artificial general intelligence. Compression is a technique that reduces the computational load on hardware by compacting AI models.

Leveraging their expertise in AI compression and AI inference-dedicated hardware technology, the two companies will optimize and compress various small language models for Rebellions' NPU.

Through this collaboration, SqueezeBits aims to establish a technical foundation for compressing generative AI models, while Rebellions plans to expand its sales channels for NPUs designed for generative AI.

Write to Joo-Wan Kim at kjwan@hankyung.com