Last Updated: 8 May 2025
Fastino has secured a $17.5 million seed round to grow its portfolio of Task-Specific Language Models (TLMs), purpose-built alternatives to general-purpose large language models (LLMs). The round was led by Khosla Ventures, known for being an early investor in OpenAI. Other participants included Valor Equity Partners, Dropbox Ventures, former Docker CEO Scott Johnston, and the co-founders of Weights & Biases, Lukas Biewald and Shawn Lewis.
This latest funding brings Fastino’s total capital raised to nearly $25 million, following a $7 million pre-seed round led by Insight Partners and M12, Microsoft’s venture arm, in late 2024. The company says it will use the new funds to scale its research team and continue building efficient, task-optimized AI tools.
Fastino is positioning its TLMs as a faster, more cost-effective solution to traditional LLMs, which are typically trained on massive datasets for broad, general usage. The founders, Ash Lewis and George Hurn-Maloney, created Fastino after seeing the inefficiencies of generalist LLMs firsthand at their previous startup.
Instead of paying for massive infrastructure to power models that handle every type of request, Fastino builds models that are tailored to specific enterprise tasks like summarization, text-to-JSON conversion, function calling, and PII redaction. These lightweight models aim to deliver higher performance on individual use cases—without overwhelming budgets.
"Large enterprises using frontier models typically only care about performance on a narrow set of tasks. Fastino’s tech allows enterprises to create a model with better-than-frontier model performance for just the set of tasks you care about."
Jon Chu, Partner at Khosla Ventures
To make its offering accessible, Fastino is introducing flat-rate monthly pricing and the industry’s first free API tier for its models. Developers can make up to 10,000 requests per month without cost, running entirely on CPUs to minimize energy use.
The company plans to use the funding to broaden this accessibility, aiming to make high-speed, low-cost models available to developers around the world. A portion of the funds will go toward hiring more researchers to fine-tune performance and expand the list of supported tasks.
Founded by Ash Lewis and George Hurn-Maloney, Fastino’s core team includes researchers from Stanford, CMU, Apple Intelligence, and Google DeepMind. Their models have been trained without the industry’s go-to high-end infrastructure — no H100s, and under $100K in compute costs — and still outperform generalist LLMs in task-specific benchmarks.
What sets Fastino apart is not just performance, but portability. The models are designed to run on CPUs and low-end GPUs, making them deployable on-prem, at the edge, or within a company’s private cloud environment. This enables greater compliance, reduced latency, and stronger control over data flows.
"AI developers don't need an LLM trained on trillions of irrelevant data points – they need the right model for their task."
George Hurn-Maloney, COO and co-founder of Fastino
Fastino’s funding journey started with a $7M pre-seed round backed by Insight Partners and M12. Its latest seed round brought in new names, including Valor Equity Partners, Dropbox Ventures, and AI industry insiders like Lukas Biewald and Shawn Lewis.
These investors join Fastino in betting that the next generation of AI won’t rely on size alone, but on specificity and speed. With its TLM architecture and flat pricing, Fastino is hoping to drive wider adoption of AI in sectors like finance, healthcare, and e-commerce.
