Nosana is a decentralized computing network for AI inference workloads, e.g., enabling trained AI models such as LLama 2 and Stable Diffusion to operate at scale. As of March 2024, network is in its Incentivized Test Grid phase, a multi-phase test network before the planned mainnet launch. Upon mainnet launch, Nosana will provide a distributed marketplace where users can access GPU computing resources to power AI workloads. Similarly, infrastructure providers can supply computing resources to the network and fulfill inference requests from individual users/developers for a fee.
Nosana was launched in 2021 by co-founders Jesse Eisses and Sjoerd Kijkstra to provide a “decentralized automation provider…for devs running CI/CD pipelines.” In October 2023, Nosana pivoted the protocol’s focus from CI/CD pipelines to AI workload jobs.
In November 2023, Nosana launched the first phase of its Incentivized Public Test Grid. Phase 1 concluded on Jan. 31, 2024, with 442 nodes participating.
In January 2024, Nosana reduced its staking emissions rate by ~30.00%, citing a goal “to strike a balance between rewarding our long-term supporters and ensuring the overall health and stability of the Nosana ecosystem.”
In February 2024, Render Network completed a successful governance vote to introduce Nosana as its fourth Compute Client. The proposal explains that Nosana will bridge the AI inference community and Render's GPU network by helping address GPU shortages, the idle compute issue, and high public cloud pricing.
Nosana’s Testnet is live and fully operational, with ~202,500 inferences completed. The network consists of ~96 online GPU nodes running ~78 inferences. Phase 2 of the Incentivized Testnet is expected to begin in Q2 2024, and users can join the waitlist here.
According to the project’s roadmap, Nosana will be launched on mainnet around August 2024. The roadmap details additional plans according to the following stages: