Callosum raises $100M for faster AI inference

Callosum raised $100 million for Tailored Inference, a service that breaks multi-step AI work into blocks and routes each block to the model and accelerator best suited to it; the company says some tasks run 3.7× faster than GPT-5.6 Luna, according to SiliconANGLE’s funding report. The round follows a $10.25 million raise in February, and the platform supports accelerators from more than a half-dozen chipmakers at launch. Builders should treat the claim as an infrastructure option to benchmark—not a promise—alongside the archive’s model-routing cost analysis, because the published comparison is task-specific.