Fastino

Fastino (fastino.ai) is a developer-first AI company operating from Palo Alto, California, that provides high-performance, task-optimized language models. Founded in 2024 by CEO Ash Lewis and COO George Hurn-Maloney, the company emerged from stealth in November of that year with a mission to solve the inefficiencies of large, general-purpose AI models for specific enterprise tasks. Lewis's journey to creating Fastino was sparked by his previous startup, DevGPT, where the high cost of using generalist LLM APIs proved to be economically unsustainable, highlighting a critical need for more efficient, purpose-built solutions.

The company has secured significant financial backing, starting with a $7 million pre-seed round in November 2024, led by Insight Partners and Microsoft's M12, with notable participation from NEA and others. This was followed by a $17.5 million seed round in May 2025, led by Khosla Ventures, bringing Fastino's total funding to approximately $25 million. This capital is being used to expand its research and development team, which includes talent from Google DeepMind, Stanford, Carnegie Mellon, and Apple Intelligence.

Fastino's core offering is a suite of what it calls Task-Specific Language Models (TLMs). Unlike conventional large language models (LLMs) that require expensive GPU clusters, Fastino's models are engineered to run on CPUs or NPUs with near-instant inference, reportedly making them up to 99 times faster on specific tasks. This unique architecture, which blends traditional transformer concepts with proprietary breakthroughs, allows for training on low-end gaming GPUs for under $100,000. The business targets enterprise AI developers who require accuracy, speed, and security for production workloads. Its business model deviates from the industry-standard per-token pricing by offering a flat monthly subscription, which includes a free tier for developers to integrate the models into their workflows.

The TLM API provides models for a range of specific enterprise functions, including summarization, function calling for agentic systems, converting unstructured text to JSON, and PII redaction. These models are designed for high accuracy in tasks like document parsing, real-time search intelligence, and ensuring content compliance. For enterprise clients, Fastino offers flexible deployment options, allowing the models to be run within a customer's Virtual Private Cloud (VPC), on-premise, or at the edge, ensuring data control and security.

Keywords: Task-Specific Language Models, TLM, enterprise AI, CPU inference, AI model optimization, low-latency AI, foundation models, Ash Lewis, George Hurn-Maloney, Khosla Ventures, Microsoft M12, developer AI tools, flat-rate AI, PII redaction, function calling, text to JSON, AI for developers, efficient AI, secure AI deployment, on-premise AI, AI model architecture