AI research lab
121-labs is an AI research lab building automated research systems - AI that studies, designs, and improves AI.
Looking beyond today's language models toward brain-inspired architectures with fundamentally better scaling laws for compute and data.
The thesis
The next leap in AI won't come from scaling further. It will come from rethinking the architecture.
Progress in AI is still bottlenecked by how much compute and data each new capability costs. The next leap won't come from scaling the same architectures further, but from rethinking them - guided by the most efficient general intelligence we know of: the human brain.
121-labs builds systems where AI drives its own research loop - proposing, testing, and refining architectures faster than human-only teams can. Our aim is general architectures that learn more from less, across modalities, not just larger language models.
Research
The work
01
Brain-inspired architectures
Designing models informed by cognition and neuroscience - sparsity, modularity, memory, and efficient knowlegde representation.
02
Better scaling laws
Searching for architectures whose performance improves more steeply with compute and data, lowering the cost of every future capability.
03
Automated AI research
Closed-loop systems where AI proposes, runs, and evaluates its own experiments, compressing research cycles from months to days.
04
Beyond foundation models
Exploring general architectures across modalities and tasks - language is one way of communication, but not the definition of intelligence.
Contact
Get in touch
Interested in 121-labs' research, collaboration, or discussions? Reach out.