Spaces:
Running
title: README
emoji: 📊
colorFrom: indigo
colorTo: indigo
sdk: static
pinned: true
🌱 turing.garden
AI for the masses.
turing.garden is an AI research and engineering company focused on creating accessible, efficient, and practical language models.
We believe that in today and tomorrow's landscape, LLMs are becoming increasingly:
- expensive (both to train and to compute);
- power hungry;
- and inaccessible to ordinary people.
Our mission
We aim to make advanced AI available to everyone by creating models that are:
- affordable to deploy and use;
- optimised for accuracy and speed;
- more energy-efficient;
- suitable for consumer hardware;
- and built with safety in mind.
What we build
Foundation models
General-purpose LLMs and SLMs trained for reasoning, understanding, and knowledge
Specialist models
Instead of forcing one model to excel at every domain, we develop specialist models for specific domains, such as:
- software engineering;
- biology, physics, chemistry;
- and general reasoning, etc.
Distilled Models
Our larger models are distilled into smaller variants that only contain millions of parameters, making them suitable for:
- mobile devices;
- resource-constrained devices;
- and more.
Research
Our work spans research and development, including:
- alignment;
- quantisation;
- efficient training and inference;
- agentic AI;
- and distillation, etc.
Open models
We support open AI where it benefits both researchers and developers.
Most models, datasets, etc, will be released under OpenRAIL.
Certain proprietary models and research may remain closed where appropriate.
Targets
We design for:
- consumer-grade hardware;
- workstations;
- and enterprise servers.
We believe AI should not require hyperscale infrastructure whenever possible.
Safety
The brain is one of the most impressive structures in the entire universe.
AI should augment human thinking and should not act as a replacement.
Our research is guided by:
- safety;
- transparency;
- and accessibility.
We do not intend for our models to facilitate:
- warfare;
- autonomous harmful systems;
- the replacement of human critical thinking.
Beyond AI
We also create develop other software!
Check us out at our website.
Philosophy
Bigger AI is not necessarily better AI. Success is measured by impact.
Through research, we believe language models can become more useful, efficient, and accessible for everyone.
Coming Soon
- Foundation language models
- Domain-specific specialist models
- Distilled lightweight models
- Technical reports
- Research publications
- Benchmarks
- Training tools
- Community resources