Fastino Raises 17.5 M To Launch TLMS That Do One Job Really Well

“AI is most valuable when it’s specialised, fast, and deployed exactly where it’s needed,” that’s the idea behind Fastino.

Fastino's journey began with a realisation that general-purpose models, while versatile, often come with high costs and inefficiencies. Drawing from experiences in their previous ventures, co-founders Ash Lewis and George Hurn-Maloney recognised the need for models tailored to particular tasks. This insight led to the development of TLMs or Task-Specific Target Models which are optimised for tasks like summarisation, function calling, text-to-JSON conversion, PII redaction, text classification, and profanity censoring.

Clearing this mission behind the formation of this startup, Lewis stated, “We started this company after our last startup went viral and our infrastructure costs went through the roof. At one point, we were spending more on language models than on our entire team.”

Smaller Models Are Winning

What sets Fastino apart is its innovative approach to model training. Unlike traditional methods that rely on expensive hardware, Fastino's models were trained on Nvidia’s low-end gaming GPUs, costing less than $100,000 in total. Developed by in-house AI researchers from Google DeepMind, Stanford, Carnegie Mellon, and Apple Intelligence, these models deliver performance that is 99 times faster than standard large language models (LLMs), making them both cost-effective and scalable.