Fastino trains AI models on cheap gaming GPUs and just raised $17.5M led by Khosla

Tech giants like to have AI models of trillion parameter that require huge and expensive GPU clusters. But Fastino follows another approach.

The Startup-based Palo Alto says it has invented a new type of AI model architecture that is deliberately small and task-specific. The models are so small that they have been trained with low-end gaming GPUs worth less than $ 100,000 in total, Fastino says.

The method attracts attention. Fastino has $ 17.5 million in seed financing led by Khosla Ventures, famous OpenAi’s first venture investor, confirmed, Fastino tells WAN exclusively.

This brings the total financing of the startup to almost $ 25 million. Last November it collected $ 7 million in a pre-seed round led by Microsoft’s VC Arm M12 and Insight Partners.

“Our models are faster, more accurate and cost a fraction to train while it performs better than flagship models on specific tasks,” says Ash Lewis, CEO and co-founder of Fastino.

Fastino has built a series of small models that it sells to customers. Each model focuses on a specific task that a company needs, such as editing sensitive data or summarizing business documents.

Fastino does not yet reveal early statistics or users, but says that performance is surprising early users. For example, because they are so small, the models can deliver a whole reaction in one token, Lewis told WAN and shows off the technology that immediately gives a detailed answer in Milliseconden.

It is still a bit early to say whether Fastino’s approach will start. The Enterprise AI space is busy, with companies such as Cohere and Databricks that also recommend AI that excels in certain tasks. And the Enterprise-oriented SATA model makers, including anthropic and Mistral, also offer small models. It is also no secret that the future of generative AI for Enterprise is probably in smaller, more focused language models.

Time can tell, but an early voice of trust from Khosla certainly doesn’t hurt. For now, Fastino says it is aimed at building an advanced AI team. It is aimed at researchers at Top AI Labs that are not obsessed with building the largest model or beating the benchmarks.

“Our recruitment strategy is very focused on researchers who may have a contrary thinking process for how language models are being built now,” says Lewis.