The Living
Language Model.
A hybrid AI system with 800,000 living human neurons in the loop. Every response passes through real biological tissue on Cortical Labs' CL1 platform before it reaches you.
Access is constrained by biology, not compute.
The problem
Silicon AI is
converging.
Biology diverges.
Frontier models keep getting better, but they are still variations on the same silicon substrate: larger datasets, larger clusters, larger optimization loops.
Evolution spent 500 million years optimising the neuron. We stopped trying to copy it and started using it.
No approximation
Language models predict the next token from statistical patterns across massive datasets. BioLLM routes your query through actual synaptic connections formed through real biological learning — not a simulation of them.
Weights that change
Silicon weights are frozen at training time. Our neurons form new connections with every session — the same Hebbian mechanism your brain uses right now.
The original network
The most energy-efficient neural network ever built is biological. We wire it into the language stack directly instead of building another silicon replica of it.
Technology
Four things no
GPU can do.
These are not features that ship in a product update. They are properties of biology — present from day one because the substrate itself is alive.
800,000 cultured neurons
Grown from human cell lines on Cortical Labs' CL1 chip. Not a model of neurons — actual neurons, maintaining electrophysiological activity through every conversation you have.
biOS: language to biology
The biological OS converts your input into electrical stimuli. The culture responds. Those spike patterns are decoded and merged with the language model output before anything reaches you.
Hybrid architecture
The language model handles syntax, structure, and retrieval. The biological layer adds physical neural dynamics — spike timing, population coding, and substrate behavior that shifts across interactions.
Synaptic plasticity
Neurons that fire together wire together. Hebbian learning physically changes the culture across a session. The system you end a conversation with is not the same one you started with.
Platform
Hardware
with a pulse.
The CL1 is the first code-deployable biological computer. It keeps neurons alive in a sealed environment — temperature-controlled, fed, and monitored continuously — for up to six months per culture.
biOS is the layer between language and biology. Your query arrives as an electrical stimulus. The culture responds with spike patterns. Those patterns come back as meaning.
Published context
The claims on this page should trace back to verifiable infrastructure, platform materials, or public research notes.
Platform partner
Cortical Labs and the CL1 biological computer
The substrate behind BioLLM is the CL1 platform from Cortical Labs. The hardware, biOS layer, and published platform material are the primary external references for how the biological side works.
Visit Cortical LabsMethodology
Research notes and implementation updates
We publish product notes, technical writeups, and methodology updates as the system evolves. The public record should become more concrete over time, not less.
Read the updatesBioethics
The biology
is documented.
Working with biological material means every claim has to be documented carefully. We prefer public methodology, explicit constraints, and outside scrutiny over dramatic language that cannot be verified.
Every neuron originates from an established, commercially available human cell line — not from living individuals. No extraction, no donors, no ambiguity.
800,000 dissociated neurons are not a conscious system — the cortical architecture, thalamic loops, and global workspace implicated in sentience are entirely absent. We hold that position with confidence, not dismissal. The uncertainty is taken seriously at every level.
We are building external bioethics review into the program and publishing methodology, constraints, and ethics notes openly as the work matures. Nothing about the ethical layer should depend on secrecy.
No conversation context persists between users. Neural activity from one session is inaccessible to another. Queries are encrypted in transit and not retained after processing.
Who's building this
A product company,
not a thought experiment.
The landing page should not ask for trust without context. This system is being built deliberately, in public where possible, and with a clear boundary between what the platform can prove today and what still needs more evidence.
Company
Founder-led and engineering-first
BioLLM is being built as a focused product and research effort at the intersection of AI systems, biological computing, and real-world interfaces. The goal is not to make louder claims than the field can support. The goal is to build something observable, testable, and worth using.
Platform
Built on public infrastructure and public context
The biological layer runs on Cortical Labs' CL1 platform and biOS stack. Our responsibility is the application layer: product experience, system integration, telemetry, and how to expose this technology honestly to users, researchers, and partners.
Questions
The ones people actually ask.
Yes — approximately 800,000 of them. They are cultured from established human cell lines on Cortical Labs' CL1 chip, housed in a life-support environment that maintains temperature, nutrients, and waste removal. They are not harvested from individuals.
SBI stands for silicon-based intelligence — conventional transformer models running on GPUs. A hybrid pairs that with biological neurons, which handle certain types of associative and contextual reasoning. Your query passes through both systems; the outputs are weighted and merged before being returned to you.
Up to six months on the current CL1 platform. The chip maintains a sealed microenvironment. When a culture reaches end-of-life it is retired, a new one is brought online, and the transition is seamless from your perspective.
biOS is Cortical Labs' biological intelligence operating system — the software layer that translates incoming stimuli into patterns the neurons respond to and interprets their electrophysiological activity back into structured data. Think of it as the driver between language and biology.
The neurons have no capacity for consciousness or suffering — they are dissociated cells, not an intact brain. We do not claim sentience, and we aim to publish methodology and ethics notes openly as the program matures. The Cortical Labs ethics material is part of the public context for this work.
API access is on the roadmap for Q3 2026, available first to Research tier partners. If you are building on it commercially, contact us now — we are taking on design partners before the public launch.
Access
Open beta now.
Priority access comes next.
Public chat is live behind authentication. Higher-capacity access for research teams, commercial partners, and future tiers is still limited by biology — not server headcount.
Authenticated public access
Open Beta
Use the product now while capacity remains available.
- Authenticated chat access
- Live biological responses while capacity is available
- Session-level neural telemetry
- Product and research updates
Labs and institutions
Research
Priority access for teams running experiments or building on the platform.
- Priority access to live capacity
- API access — Q3 2026
- Expanded neural telemetry
- Design-partner collaboration
- Direct Slack channel with our team
Dedicated infrastructure
Enterprise
Reserved capacity and deployment support for commercial teams.
- Reserved CL1 capacity
- Private, air-gapped deployment option
- Custom integration support
- 99.9% uptime SLA
- Named account engineer
Early access
Open beta is live.
Capacity is still scarce.
Access is constrained by biology, not compute. Use the authenticated beta now, or join the access list for priority capacity as more cultures come online.
CL1 Platform · Cortical Labs · biOS v1