OneBit
We train a new species of AI model, small by design rather than shrunk afterwards. It holds the most intelligence per byte, on the chips you already own, not in a data center.
On any device with a processor: phones, laptops, servers, cameras, drones, robots, satellites.
Runs on:
Any CPUAny CPUThe next Moore's law
For fifty years, more intelligence meant more transistors. The next curve is how much intelligence fits in a byte.
10×
A tenth of the space, the same intelligence
The mostintelligenceper byte.
OneBit
We train a new species of AI model, small by design rather than shrunk afterwards. It holds the most intelligence per byte, on the chips you already own, not in a data center.
On any device with a processor: phones, laptops, servers, cameras, drones, robots, satellites.
Runs on:
Any CPUAny CPUThe next Moore's law
For fifty years, more intelligence meant more transistors. The next curve is how much intelligence fits in a byte.
10×
A tenth of the space, the same intelligence
The mostintelligenceper byte.
Servers
Most of what a company sends to an LLM API is small, repeated work: sorting tickets, pulling out a name or a date, deciding where a request goes, summarizing a thread. A small model on a CPU in the building does that part.
- Classification, extraction, routing
- Runs on the CPU you already have
- No price per request
Consumer devices
A voice assistant that works without internet. Live transcription and translation. Search through your own files and photos, with nothing uploaded to anyone.
- Phones, laptops, TVs, cars
- Works with no internet
- Nothing leaves the device
Cameras, drones, robots
Find defects on a production line. Spot disease in a crop. Read a gauge or a serial number. The camera does the work and sends only the result.
- Inspection, sorting, navigation
- Runs on the camera itself
- Only results are sent
Wearables and sensors
Detect a wake word. Track heart rate and sleep. Let a toy recognize what it is looking at. The model is small enough to run next to the sensor.
- Watches, earbuds, toys
- Megabytes, not gigabytes
- All day on one charge
Microcontrollers
Read a meter. Detect a machine that is starting to fail. Notice when a package has been damaged. One battery lasts for years.
- Monitoring and fault detection
- Years on one battery
- No GPU, no cloud
What it changes
Nothing is uploaded. The model reads the data where it is produced, on the device that produced it. There is no copy of it anywhere else.
No connection needed. The device keeps working in a tunnel, in a field, on a plane, and on a factory floor with no signal. It never waits for a server to answer.
No inference bill. Running the model costs what the device already costs to run. There is no price per request that grows with how much the product is used.
No GPU. A CPU is enough, and at the small end a microcontroller. The hardware is already inside the product.
Intelligenceat theedge.
The four reasons intelligence can leave the data center.
What a model knows is a map
A model is billions of connections. Each one strengthens a signal, weakens it, or leaves it alone. What the model knows is that map: which connections exist and which way they push. The exact strength is stored to a precision the map never uses.
- Billions of connections
- Strengthen, weaken, or leave alone
- The map is the knowledge
What a model knows is a map
A model is billions of connections. Each one strengthens a signal, weakens it, or leaves it alone. What the model knows is that map: which connections exist and which way they push. The exact strength is stored to a precision the map never uses.
- Billions of connections
- Strengthen, weaken, or leave alone
- The map is the knowledge
Most of the detail is waste
Keep the map and drop the precision, and the model keeps working. The same model takes about a tenth of the space, and its arithmetic turns from multiplication into addition.
- A tenth of the space
- Addition instead of multiplication
- Nothing lost that matters
Most of the detail is waste
Keep the map and drop the precision, and the model keeps working. The same model takes about a tenth of the space, and its arithmetic turns from multiplication into addition.
- A tenth of the space
- Addition instead of multiplication
- Nothing lost that matters
Ordinary chips are enough
Addition is what ordinary processors are good at. A model in this form fits in the memory of a laptop, a server or a phone, and runs on the chip that is already there. No data center, no internet. A GPU is optional, for even faster inference.
- Any CPU
- Works offline
- GPU optional
Ordinary chips are enough
Addition is what ordinary processors are good at. A model in this form fits in the memory of a laptop, a server or a phone, and runs on the chip that is already there. No data center, no internet. A GPU is optional, for even faster inference.
- Any CPU
- Works offline
- GPU optional
Anything with a chip can think
When intelligence is this small it stops being a service you rent and becomes a part you build in. Cameras, drones, robots, satellites, factory machines: anything with a chip can carry its own.
- Built in, not rented
- Cameras, drones, robots, satellites
- Every chip, eventually
Anything with a chip can think
When intelligence is this small it stops being a service you rent and becomes a part you build in. Cameras, drones, robots, satellites, factory machines: anything with a chip can carry its own.
- Built in, not rented
- Cameras, drones, robots, satellites
- Every chip, eventually
Stay updated with our next releases.
No spam. Unsubscribe any time.
No spam. Unsubscribe any time.