Fastino trains AI models on cheap gaming GPUs and ...

Tech giants prefer to boast about trillion-parameter AI fashions that require huge and costly GPU clusters. However Fastino is taking a special method.

The Palo Alto-based startup says it has invented a brand new sort of AI mannequin structure that’s deliberately small and task-specific. The fashions are so small they’re skilled with low-end gaming GPUs value lower than $100,000 in complete, Fastino says.

The tactic is attracting consideration. Fastino has secured $17.5 million in seed funding led by Khosla Ventures, famously OpenAI’s first enterprise investor, Fastino solely tells TechCrunch. 

This brings the startup’s complete funding to almost $25 million. It raised $7 million final November in a pre-seed spherical led by Microsoft’s VC arm M12 and Perception Companions.

“Our fashions are sooner, extra correct, and value a fraction to coach whereas outperforming flagship fashions on particular duties,” says Ash Lewis, Fastino’s CEO and co-founder.

Fastino has constructed a collection of small fashions that it sells to enterprise clients. Every mannequin focuses on a particular process an organization may want, like redacting delicate information or summarizing company paperwork.

Fastino isn’t disclosing early metrics or customers but, however says its efficiency is wowing early customers. For instance, as a result of they’re so small, its fashions can ship a complete response in a single token, Lewis advised TechCrunch, displaying off the tech giving an in depth reply without delay in milliseconds. 

It’s nonetheless a bit early to inform if Fastino’s method will catch on. The enterprise AI area is crowded, with firms like Cohere and Databricks additionally touting AI that excels at sure duties. And the enterprise-focused SATA mannequin makers, together with Anthropic and Mistral, additionally provide small fashions. It’s additionally no secret that the way forward for generative AI for enterprise is likely in smaller, more focused language models.

Time could inform, however an early vote of confidence from Khosla definitely doesn’t damage. For now, Fastino says it’s targeted on constructing a cutting-edge AI staff. It’s concentrating on researchers at prime AI labs who aren’t obsessive about constructing the largest mannequin or beating the benchmarks.

“Our hiring technique may be very a lot targeted on researchers that perhaps have a contrarian thought course of to how language fashions are being constructed proper now,” Lewis says.