Pluralis Research

Sign in to add this project to your watchlist

Overview

Tickers
None recorded
Sector
machine-learning infrastructure
Lifecycle stage
Development
Pre-TGE classification
Uncertain
Token status
Unknown
Last updated

Research lab working on decentralized AI training, an approach it calls Protocol Learning, in which models are trained across many participants' machines over the internet and owned by the collective. No token has been announced.

Token status summary

Unknown

Synthesis

No synthesis has been written yet.

Product

Pluralis Research's documentation describes Agora as a system for collectively pretraining large language models over the internet, across globally distributed, heterogeneous and unreliable hardware including home devices.

Supported

Project announcement

Evidence (2): 2 supporting
  • supports: Pluralis Research
    Agora is a system for collectively pretraining large language models over the internet.
  • supports: Pluralis Research
    Agora can access and train across globally distributed, heterogeneous and unreliable hardware, including local devices connected to home internet

In March 2025 Fortune reported that Pluralis did not yet have a product, so investors were mainly betting on the founder and his team of researchers.

Supported

Evidence (1): 1 supporting; 1 needs review
  • supports: Fortune
    For now Pluralis doesn’t have a product, meaning investors are primarily betting on Long and the veteran team of AI researchers he’s assembled.

    Needs review

Pluralis Research's March 2025 press release describes Protocol Learning as training large AI models over open, permissionless networks of globally distributed compute, with models that live within the protocol and are trainable and usable but unextractable.

Supported

Project announcement

Evidence (2): 2 supporting

Pluralis Research's website describes it as a research lab focused on decentralized AI, where models are developed, trained and served across global networks of many participants and owned by the collective, an approach it calls Protocol Learning.

Supported

Project announcement

Evidence (1): 1 supporting
  • supports: Pluralis Research
    Pluralis is a research lab focused on decentralized AI. We believe the best path is where the models are developed, trained and served across global networks of many participants and owned by the collective, an approach we call Protocol Learning.
Back to sections

Team

The current team size and leadership beyond the founder are unknown. The most recent dated count is from March 2025, and the 2026 research papers list roughly ten authors each.

Unknown

Evidence (1): 1 context; 1 needs review

In March 2025 Fortune named Alexander Long as Pluralis's founder and CEO, with a computer science doctorate and more than three years at Amazon as an AI engineer, and reported that he had assembled seven other computer scientists with doctorates or postdoctoral research experience.

Supported

Evidence (3): 3 supporting; 3 need review
  • supports: Fortune
    The Pluralis founder has a doctorate in computer science and worked at Amazon for more than three years as an AI engineer.

    Needs review

  • supports: Fortune
    Alexander Long is the founder and CEO of Pluralis Research

    Needs review

  • supports: Fortune
    He’s assembled seven other computer scientists, all with doctorates or stints as postdoctoral researchers

    Needs review

Pluralis Research's March 2025 press release states that it was founded in early 2024, is based in Melbourne, Australia, and is a team of ex-FAANG research scientists led by Alexander Long.

Supported

Project announcement

Evidence (2): 2 supporting
Back to sections

Funding

Any financing after the March 2025 round is not yet verifiable. No later round was found in the sources checked.

Not yet verifiable

No evidence recorded.

On 19 March 2025 Pluralis Research announced a $7.6 million seed round co-led by USV and CoinFund, with Topology, Variant, Eden Block and Bodhi Ventures and angels Balaji Srinivasan and Clem Delangue. Fortune calls it pre-seed and seed funding.

Supported

Evidence (3): 3 supporting
  • supports: Fortune
    Topology, Variant, Eden Block, and Bodhi Ventures also participated in the round, along with prominent crypto investor Balaji Srinivasan and Clem Delangue, cofounder of the popular AI platform HuggingFace.
  • supports: Fortune
    On Wednesday, Pluralis announced that it had secured $7.6 million in pre-seed and seed funding, led by the venture capital firms CoinFund and Union Square Ventures.
  • supports: Pluralis Research via GlobeNewswire
    today announced its $7.6M seed funding round co-led by USV and CoinFund with participation from Topology, Variant, Eden Block, and Bodhi Ventures.
Back to sections

Technical progress

Pluralis Research's 2026 technical report on Agora describes Pluralis-8B, an 8.6B-parameter open pretraining run trained on 500B tokens across 330 contributor nodes.

Supported

Project announcement

Evidence (1): 1 supporting
  • supports: Pluralis Research
    The report describes Pluralis-8B, an 8.6B-parameter open pretraining run trained on 500B tokens across 330 contributor nodes.

Pluralis Research's documentation lists three training runs: Node0-7.5B (September 2025, 36.6B tokens), Pluralis-8B (May 2026, 531.8B tokens) and Pluralis-13B (September 2026, 62.4B tokens, described as a systems test of Agora's upgraded architecture).

Supported

Project announcement

Evidence (3): 3 supporting

In a post about its multi-party training stack (January 2026) Pluralis Research called Node0 version 0.0.1 and said it has a long way to go until decentralized training is competitive with the centralized alternative.

Supported

Project announcement

Evidence (1): 1 supporting
  • supports: Pluralis Research
    Node-0 is the v0.0.1, and we have a long way to go until decentralized training is competitive with the centralized alternative.

Pluralis Research's website lists peer-reviewed papers including Subspace Networks (NeurIPS 2025), which it says trained an 8B LLaMA model on par with centralized training with blocks split across four locations over standard internet links.

Supported

Project announcement

Evidence (1): 1 supporting
  • supports: Pluralis Research
    This work demonstrates that model-parallel training over low-bandwidth networks is possible, training an 8B LLaMA model on par with centralized training while transformer blocks are split across four locations connected only by standard internet links.
Back to sections

Adoption

Pluralis Research's documentation says its training runs bring together consumer GPUs at home, instances from major cloud providers and small-scale operators on GPU marketplaces, and its report counts 330 contributor nodes for Pluralis-8B.

Supported

Project announcement

Evidence (2): 2 supporting
  • supports: Pluralis Research
    trained on 500B tokens across 330 contributor nodes
  • supports: Pluralis Research
    Each run brings together consumer GPUs at home, instances from major cloud providers and small-scale operators on GPU marketplaces.

Independent confirmation of contributor numbers and of any use of Pluralis models beyond its own runs is not yet verifiable. The public dashboard returned a service-unavailable page when fetched on 10 November 2026.

Not yet verifiable

No evidence recorded.

Back to sections

Competitors

Fortune (March 2025) named Gensyn and Prime Intellect among firms exploring decentralized AI training, and quoted Alexander Long saying Pluralis's approach differs because most attempts require every computer to download the entire model.

Supported

Evidence (2): 2 supporting
  • supports: Fortune
    Pluralis’s approach is different, Long said. Most attempts at training AI through a network of decentralized computers require those computers to download the entire model.
  • supports: Fortune
    Gensyn, Prime Intellect, and a handful of firms are already exploring the possibility.

How Pluralis compares with competing decentralized-training projects today is unknown. It has not been researched beyond the March 2025 Fortune article.

Unknown

No evidence recorded.

Back to sections

Token status

Whether Pluralis has decided to launch a token is unknown. No token, ticker, token sale, airdrop or contract address appears in its website, documentation, blog or GitHub pages read on 10 November 2026, and the Agora points carry no financial reward. A data aggregator lists a token allocation for Pluralis, but its page could not be read.

Unknown

Evidence (2): 2 context; 2 need review
  • context: Pluralis Research

    Needs review

  • context: Fortune
    The raise was for equity, with a warrant for future cryptocurrency—if Pluralis decides to launch one, Alexander Long, founder and CEO, told Fortune.

    Needs review

Pluralis Research's founder-authored article on Protocol Learning says incentives would take the form of partial model ownership allocated in proportion to training contribution. It does not describe a token.

Supported

Project announcement

Evidence (1): 1 supporting
  • supports: Pluralis Research
    Protocol Learning introduces such incentives in the form of partial model ownership allocated proportional to training contribution.

Pluralis Research's documentation says its Agora leaderboard points reflect contribution only and that no financial reward is tied to a score at this stage.

Supported

Project announcement

Evidence (1): 1 supporting; 1 needs review
  • supports: Pluralis Research
    No financial incentive yet Points reflect contribution only. There is no financial reward tied to your score at this stage.

    Needs review

Investor USV's March 2025 post says Pluralis's finished models remain within the protocol, are protocol-owned, and have ongoing usage revenue distributed back to contributors. This is the investor's description, not a Pluralis token announcement.

Unverified

Evidence (1): 1 supporting
  • supports: Union Square Ventures (investor)
    In other words, the models are “protocol-owned”, with the ongoing usage revenue distributed back to all contributors.

In March 2025 Fortune reported that Pluralis's seed round was for equity with a warrant for future cryptocurrency if Pluralis decides to launch one, according to its founder and CEO.

Supported

Evidence (1): 1 supporting; 1 needs review
  • supports: Fortune
    The raise was for equity, with a warrant for future cryptocurrency—if Pluralis decides to launch one, Alexander Long, founder and CEO, told Fortune.

    Needs review

Back to sections

Risks

Pluralis Research describes its decentralized training as early, saying Node0 is version 0.0.1 and has a long way to go, and Fortune reported in March 2025 that it had no product.

Supported

Evidence (2): 2 supporting; 1 needs review
  • supports: Pluralis Research
    Node-0 is the v0.0.1, and we have a long way to go until decentralized training is competitive with the centralized alternative.
  • supports: Fortune
    For now Pluralis doesn’t have a product, meaning investors are primarily betting on Long and the veteran team of AI researchers he’s assembled.

    Needs review

Other risks, including security of its protocol, legal exposure from a future token or warrants, and dependence on a small research team, have not been researched for this dossier and are unknown.

Unknown

No evidence recorded.

Back to sections

Sources

Back to sections
Back to discovery

Nothing on this site is financial advice.