What we believe
Frontier intelligence, designed from the ground up to be efficient.
"...committed to 30 gigawatts of compute..."
"...cost of ownership over $1.4 trillion..."
"...1 gigawatt per week of new capacity..."
"...need 10 billion GPUs..."
"...three orders of magnitude off..."
"...First Gigawatt training cluster..."
"...hundreds of gigawatts or more..."
"...600k H100 equivalents..."
"...energy consumption is the bottleneck..."
"...building the Stargate supercomputer..."
"...committed to 30 gigawatts of compute..."
"...cost of ownership over $1.4 trillion..."
"...1 gigawatt per week of new capacity..."
"...need 10 billion GPUs..."
"...three orders of magnitude off..."
Params
Higher is better · illustrative
The industry is racing to spend trillions of dollars on raw computing power. Deca takes a different path. Our Dynamoe architecture instantly figures out exactly which parts of the model are needed for a task and only turns those on. World-class intelligence shouldn't require a supercomputer to run.
When AI is cheap enough, it stops being something you just consult and becomes something that runs alongside you constantly. Agents stay active. Context stays warm. You can iterate freely instead of worrying about API costs.
Throwing billions of dollars at massive data centers isn't a strategy; it's a risk. While competitors rush to build the biggest supercomputers possible, we view efficiency as a core safety mechanism. By building smarter, more efficient models, we can push the frontier of intelligence without blindly scaling risk.
Meet the Models
Choose the perfect model for your needs. From everyday tasks to complex research.
Deca 2.5 Mini
deca-2.5-mini
Small but intelligent model for everyday tasks.
$2.50 free credits when you sign up
Deca 2.5 Pro
MOST POPULARdeca-2.5-pro
For power users who need more intelligence for less cost.
$2.50 free credits when you sign up
Deca 2.5 Ultra
BETAdeca-2.5-ultra-preview
Our bleeding-edge model for tasks requiring extreme intelligence.
$2.50 free credits when you sign up
Latest from Deca
Deca 2.5: Intelligence on Demand
Our latest models—Mini, Pro, and Ultra—prove that you don't need brute-force data centers to get world-class results. By using Dynamoe routing, Deca 2.5 breaks the trade-off between how smart a model is and how much it costs to run.
Read more →The Deca Manifesto
Our view on what AI should be, how it should behave, and why the current industry approach to safety and scaling is missing the mark. We believe the defining word for an AI model shouldn't just be "smart"—it should be "wise."
Read more →