Basic Concepts
What is inference
If you've used ChatGPT — you already know what inference is. Every time you type a question and get an answer from a neural network — that's inference. Let's delve deeper.
Simple explanation
You type a question to ChatGPT — you get an answer. This process is called inference. The neural network “thinks” and generates an answer word by word. Each such request requires a powerful GPU — a video card that can quickly process neural network computations.
Inference in the Gonka network
When someone sends a request via the Gonka API, one of the nodes processes it - a GPU server connected to the network. Currently, the network supports two models: Qwen3-235B (235 billion parameters) and Kimi K2.6 from Moonshot AI, added via the DevShards mechanism in May 2026. The host-owner of the server receives GNK for each processed request. The more powerful the GPU, the more requests the node can handle.
Why inference is the basis of GNK value
The more people use AI through Gonka — the greater the demand for GPU power. Higher demand — higher network load — higher inference price — more earnings for hosts. Unlike Bitcoin, the value of GNK is tied to the real market of AI computations, not to abstract “difficulty”.
Inference is a request to a neural network. Every time someone uses AI through Gonka, hosts earn GNK. More users = more token value.
Want to learn more?
Understand the GNK economy or start earning right now.
Read also
Technology
What is Proof of Work 2.0
Technology
Proof-of-Computation V2: how node honesty is verified
Tokenomics
How GNK is distributed: genesis vs revenue
Security
Collateral and slashing: how the network is protected
Tokenomics
Vesting: why rewards aren't immediate
Architecture
What are Transfer Agents in Gonka
Technology
DiLoCo: Distributed Model Training in Gonka
Technology
Sprint: How Gonka's Consensus Works
Basic Concepts
What is GNK: The Gonka Network Token
Basic Concepts
Epoch in Gonka: Network Time Unit
Tokenomics
Community Pool: The Gonka Ecosystem's Common Fund