In a world where tech giants flaunt trillion-parameter AI models and billion-dollar GPU clusters, Palo Alto-based startup Fastino is taking a radically different path training powerful, task-specific AI models using low-cost gaming GPUs.
Fastino has just raised $17.5 million in seed funding, led by Khosla Ventures, bringing its total funding to nearly $25 million. The company previously secured $7 million in a pre-seed round backed by Microsoft’s M12 and Insight Partners.
Unlike traditional AI models, Fastino’s architecture is intentionally small, fast, and highly specialized. These models cost a fraction to train, using less than $100,000 worth of gaming GPUs, and are already impressing early users with their blazing speed and accuracy.
CEO Ash Lewis says the company is already offering a suite of AI models to enterprise clients, focusing on specific tasks like document summarization and data redaction. The models can generate entire responses in milliseconds, delivering information in a single token.
While the enterprise AI space is competitive with players like Cohere, Databricks, and Anthropic Fastino’s efficient approach is gaining attention. The startup is currently building out a team of AI researchers who believe smaller, smarter models may be the future of generative AI.
Stay tuned for more tech breakthroughs at Jeffkom Story your source for the latest startup and AI innovation news.
