Bagel
Paris, FR · Founded 2023 · 9 known investors
Bagel Labs develops distributed training methods for diffusion models using commodity hardware across heterogeneous systems. The company builds decentralized diffusion model architectures that replace single large models with ensembles of smaller expert models, enabling training of generative models for robotics, video, and world modeling on non-homogeneous GPU infrastructure.
Also known as Bagel Labs
Founders & leadership
Bagel was founded in 2023 by Bidhan R and Bidhan R..
Investors · 9
__Twitter.png)
Also in the syndicate · 2
Funding
SEC filings, press & company announcements- Undisclosed amountseries AJun 2023
VERSO Capital (lead), Craft Lane Capital, Gaingels, Seven Seven Six
Source ↗
Source: company announcements and press reports — follow each round's link for the claim.
Company profile
researched Aug 2026Bagel Labs is a physical AI research lab working on a general world action model intended to control autonomous robots across different tasks, environments and robot types, with the stated goal of a single model that improves recursively through experience.
The company describes two core pieces of technical work. PARIS is its training architecture, in which separate components of a robotics model are trained independently and then composed into a single model; the company cites published matched-resource image- and video-generation tests in which PARIS 1.0 improved results by 24% and PARIS 2.0 by 50%, noting these are not robotics validations. WorldDiT is its compact model architecture for robot learning and control, positioned on a reported LIBERO simulation Pareto frontier trading off model size against task success, which the company notes is simulation evidence rather than real-robot validation. It also references PIE, the Paris Inference Engine.
Latest developments
The company site describes PARIS 2.0 results and the WorldDiT architecture, and lists open roles via a careers page.
▸Full profile — market position, technology, geography
Market position
Positions itself as a research lab pursuing a general world action model for robotics rather than a task- or platform-specific robotics vendor.
Technology
The company's stated technical stack centers on decentralized/compositional diffusion-style modeling. PARIS is a training architecture that lets distinct parts of a robotics model learn independently before being composed into one model, described alongside a decomposition of the diffusion objective into per-component score fields. WorldDiT is a compact model architecture for robot learning and control, evaluated on the LIBERO simulation benchmark for the trade-off between model size and task success. PIE (Paris Inference Engine) is referenced as the associated inference component.
Geography
The company site references Paris, and also names its training architecture PARIS; a Toronto, Canada reference also appears on the site.
Compiled by commissioned research from 8 cited public sources — announcements, filings, and press listed under research sources below.
Key figures
latest reportedCompany-reported or press-reported figures, each dated to when it was claimed — not independently audited.
Competitors · 5
by search overlap
Companies competing with Bagel for the same Google search keywords, organic and paid, via search-intersection analysis.
In the news
▸Research sources · 8
primary sources listed
- Bagelbagel.com · web
8 public sources were cited for this profile; the first-party ones are listed here.
Frequently asked questions
- What does Bagel do?
- Bagel Labs is a physical AI research lab building a general world action model for autonomous robot control.
- Who founded Bagel?
- Bagel was founded by Bidhan R, Bidhan R. in 2023.
- Who are Bagel's investors?
- Bagel's investors include Borderless Venture, Breed, CoinFund, Digital Currency Group, Maven 11, Gaingels, Seven Seven Six.
- Where is Bagel headquartered?
- Bagel is headquartered in Paris, FR.