Forgebyte lands $80M Series B to scale an on-device multimodal AI inference stack

AI · 6 min read

Forgebyte announced an $80 million Series B led by Granite Ridge with strategic participation from chipset vendors and mobile OEMs. The funding will accelerate development of its inference SDK that runs vision, audio, and language models efficiently on-device using compiler optimizations and quantization pipelines.

The company claims its stack delivers 3-5x speedups over existing runtimes for common multimodal workloads and supports dynamic batching, model swapping, and privacy-preserving pipelines. Forgebyte is partnering with app developers to embed the runtime into AR, accessibility, and productivity apps.

Forgebyte plans to support bare-metal builds for popular ARM and RISC-V designs and will release benchmarks and a model zoo tailored to mobile-first use cases. The infusion of capital will expand its global engineering presence and drive integrations with major machine learning frameworks.