Unlocking Collective Intelligence: The Tiny Labs Mission

Tag: General · Dec 15, 2024

Our Mission

Tiny Labs is a community-driven research laboratory spun out of the University of Cambridge, dedicated to sparking the “DeepSeek moment” for small language models. Our goal: build the world’s first sub-billion parameter language model to surpass 80% accuracy on MMLU—a benchmark typically dominated by much larger models.

Why Small Models?

We believe the next frontier in AI is not scale, but efficiency and accessibility. Small language models open up a categorically different set of opportunities: they can run seamlessly on consumer devices, enable fine-tuning on the edge, and can be trained in-house with limited computational resources. This shift empowers individuals, startups, and researchers everywhere to innovate without the need for massive infrastructure.

The Need for Change

As the demand for deployable, energy-efficient, and transparent AI grows, individual labs and companies are racing to develop their own small language models. However, siloed efforts face limits in speed, reproducibility, and impact. The current AI research ecosystem is highly centralised and inefficient, with a vast, untapped reserve of expertise scattered across GitHub, online communities, and student groups.

Our Story

We are Pietro and Richard, two Cambridge PhD grads. During our PhD, we shared an office; now we share the same passion for building the world’s best small language models. Our journey began with a simple question: what if the next leap in AI came not from more parameters, but from more people? We believe the future of AI innovation lies in harnessing the community’s collective intelligence.

Vision for a New Paradigm

Tiny Labs introduces a new paradigm: community-led development of small language models through a transparent, structured, and incentivised research platform. Our mission is to unlock the collective intelligence of the global research community by making it as easy to contribute to frontier AI as it is to contribute to open-source software.

We envision a world where anyone with a good idea—not just those with credentials or compute—can contribute to frontier research and see their work make a difference. Our approach combines a shared infrastructure, a structured experiment pipeline, and meaningful incentives to unlock the full potential of the global AI community.

Recognition, Incentives, and Open Progress

We believe that recognition and incentives are key to unlocking participation. At Tiny Labs, contributors receive financial rewards based on measurable impact, and their influence is tracked across key areas of model development. Our monthly technical report highlights key findings and progress, with all contributing members listed as co-authors—providing formal recognition and a citation-worthy record of participation.

Join Us

We invite contributors, sponsors, and AI enthusiasts to join us in building a new kind of research organisation: community-driven, focused on small models, and designed from the ground up to be open and inclusive. Now is the time to invest in small, open language models. Consumer hardware is increasingly capable of running them efficiently, public demand for transparent and auditable AI is growing, and open initiatives like Big Science and SmolLM have demonstrated what's possible—though these efforts remain fragmented. Tiny Labs offers a coordinated, scalable alternative: a platform built for real-world impact and collective innovation.