Gradient Network

Gradient Network is a decentralized AI infrastructure protocol on Solana that turns idle personal devices into a distributed compute network, challenging centralized cloud providers by building an open, community-owned alternative for AI training and inference.

What Gradient Network Does

Gradient's core thesis is straightforward: the compute needed to train and serve large language models does not have to be concentrated in hyperscaler data centers. Billions of laptops, desktops, and GPU rigs sit underutilized globally. Gradient aggregates that latent capacity into a permissionless, autonomous network — and settles the economics of that coordination on Solana.

The protocol is organized around a product framework called the Open Intelligence Stack (OIS). Rather than a single product, OIS is a layered system of primitives covering the three principal bottlenecks in distributed AI: data movement, inference, and post-training. Each layer has its own component:

• Lattica — The data motion layer, described by the team as the "Universal Data Motion Engine" and the connective tissue of the stack. Lattica handles peer-to-peer communication between distributed nodes, routing data reliably across heterogeneous devices on the public internet.

• Parallax — The inference layer, positioned as a "Sovereign AI Operating System." Released as open source in November 2025, Parallax enables developers to shard large foundation models across multiple machines and run them as a unified compute surface. It supports over 40 open-source AI models and runs on Windows, macOS, and Linux, across both NVIDIA GPUs and Apple Silicon. Internal benchmarks across 14 connected machines showed 3.6x higher throughput versus leading local hosting alternatives and 2.6x lower latency. The software provides auditable, reproducible computation and keeps data local by default.

• Echo — The orchestration and post-training layer. Echo is a distributed reinforcement learning framework that decouples the inference step from the training loop, enabling asynchronous RLHF-style workflows across a fleet of nodes without requiring synchronized clusters. Echo-2, launched February 11, 2026, reduced AI post-training costs by up to 80% compared to traditional cloud approaches — with some benchmarks showing over 10x cost improvement relative to cloud baselines.

Together, these three primitives address the question Gradient poses directly in its documentation: how do you train and serve good LLMs over the public internet on a permissionless, autonomous network?

Node Participation and Network Roles

Gradient has iterated through several participation models as the network matured.

The earliest entry point was the Sentry Node, a lightweight browser extension that monitored network connectivity and contributed bandwidth and uptime data. Sentry Node participation ran across Seasons 0 and 1, generating EXP points redeemable in a future token event. This program has since concluded.

The active participation tier is Worker Nodes, which contribute compute to inference and training tasks. Nodes are assessed through a Node Score system that weighs hardware capability, uptime, and task performance, with rewards distributed accordingly. An Edge Host pilot program represents the next stage of participation, allowing hardware operators to serve as dedicated edge compute providers within the OIS stack.

Gradient Cloud is the enterprise-facing product built on top of this infrastructure: an OpenAI-compatible API that routes requests across the distributed network, giving commercial users access to affordable AI inference without requiring them to manage the underlying node infrastructure.

Solana Integration

Gradient chose Solana as its settlement layer for a specific set of reasons: the chain's throughput and sub-second finality allow compute payments to be settled in real time as tasks complete, rather than batching settlements at the end of a session. Solana's negligible transaction fees make micro-settlements economically viable in a way that most other chains cannot match.

Team and Funding

Gradient was co-founded by Eric Yang, who serves as CEO, and Yuan Gao, who previously led growth at Helium Foundation. Gao's Helium background is directly relevant: Helium pioneered the model of community-owned wireless infrastructure rewarded with protocol tokens, and Gradient applies the same structural logic to AI compute.

In June 2025, the project closed a $10 million seed round led by Pantera Capital and Multicoin Capital, with additional participation from HSG (formerly Sequoia Capital China). The raise was one of the larger early-stage AI-DePIN rounds on Solana that year.

Token and Incentives

As of mid-2026, Gradient has not announced a Token Generation Event (TGE) date, tokenomics structure, or vesting details. The EXP points accumulated by Sentry Node participants during Seasons 0 and 1 are expected to convert to a native token, but the conversion ratio and distribution schedule remain to be announced. The project originally signaled a Q4 2025 TGE, which did not occur; the timeline shifted alongside the major product launches of Parallax (November 2025) and Echo-2 (February 2026).

No token currently trades. There are no announced audit reports as of this writing.

Ecosystem Position

Gradient occupies an increasingly competitive segment of the Solana ecosystem: decentralized AI infrastructure. It competes and overlaps with projects like io.net, which aggregates GPU compute for ML workloads. What distinguishes Gradient is its vertical integration — rather than being purely a GPU marketplace, it is building the software stack (OIS) that runs on top of the distributed hardware, including a full distributed RL framework and an inference OS. This positions the protocol closer to the application layer and gives it a more defensible product surface than commodity compute pooling.

The combination of credible founding experience (Helium DePIN), strong institutional backing, and consistent product delivery in 2025–2026 makes Gradient one of the more substantive AI-DePIN projects to have emerged in the Solana ecosystem.