We are excited to announce Datawizz’s $12.5M Seed round led by Human Capital - to further our journey of making AI efficient, economical and sustainable. Datawizz drives a >85% reduction in AI cost by replacing generic LLMs with Specialized Language Models.
Companies spent $8.4B on LLM API calls in the first half of 2025 - more than all of 2024. Adoption is undeniable; the challenge is that many applications have upside down unit economics and deliver inconsistent results. The current approach to building AI applications is broken - and we have the solution.
The current approach is to use huge general-purpose language models for every task. Your coding assistant, AI SDR, and calorie counter app, all call on the same model. This requires training ever larger models capable of “generalizing” well across a huge host of tasks. This is the AI equivalent of staffing a company only with people who have every PhD. Bonkers, and it shows up in your margins.
We present a fundamentally different approach - replacing large generic models with small, specialized models trained to do one thing extremely well. Your coding assistant might use dozens of different models depending on the programming language, task and available tools - each model specialized at a specific task.
This is not a novel approach - large players are already looking to Specialized Models as a way to optimize AI performance. However until today, transitioning to specialized models required in-house AI/ML expertise that few have access to. The onus of training, evaluating, deploying and observing these models fell to the developer and required in-depth expertise.
The Datawizz Platform
Datawizz offers a new paradigm - a fully managed, end-to-end platform that automatically identifies common requests, distills them into models 1,000s of times smaller than popular LLMs, and routes traffic to the right model every time. From a developer perspective, it’s a transparent layer. From an operational perspective, we bring >85% cost savings, 5-15x improvement in speed, and more precise and predictable results (especially when using human feedback).
On top of that - Datawizz creates specialized models you own. The dependency on LLMs from the large providers puts companies directly at their mercy. When performance degrades, APIs go down, or providers decide to nerf their models companies have little control over their AI systems. With Datawizz, you get the benefit of small models you own, control and manage. Run them on-prem, on device or in your private cloud, no data shared with anyone.
The Future with Datawizz
By 2030, ~70% of AI traffic will be served by specialized, sub-5B parameter models. We predict more than 50% of that traffic will flow through and be managed by Datawizz.
With this recent funding, we are growing our SF based team to bring small, specialized models to every company. If you are excited about this mission and want to build the next generation of AI with us, reach out at [email protected] .
Iddo & The Datawizz Team
