ezo labs builds worlds for agents.
we build the worlds agents work in. then the swarms that run them. on our models, on clusters we build.
what ezo labs does
-
01
worlds
the places agents work, built first.simulation, environment, orchestration.
-
02
agents
clusters and swarms, trained per world.mixed across model families.
-
03
superbot
agents to manage your agents.in private beta at superbot.gg.
what makes ezo labs different
the world first, not the model
an agent is only as good as its world.
clusters and swarms, not one agent
we mix model families in one team.
our own agent engine, omni
built in house, measured below.
our own inference router
each call goes to the model that fits.
we fine-tune our own models
tuned to the worlds they work in.
we build our own gpu clusters
training runs on machines we put together.
omni, measured
omni is our agent engine.here it is against the usual ones.
judged quality
sonnet 5, higher is better
time per task
sonnet 5, lower is better
tasks finished
haiku 4.5, higher is better
output tokens per task
haiku 4.5, lower is better
who it's for
- teams whose agents break in the world.
- labs that need worlds to train against.
- operators who run many agents at once.
how to work with us
- contact@ezolabs.io
key facts
- company name
- ezo labs inc
- type
- delaware corporation
- founded
- 2026
- headquarters
- wilmington, delaware, usa
- website
- ezolabs.io
- core offering
- worlds for agents, and the agents to run them
- products
- superbot, agents to manage your agents
- engine
- omni, our own
- inference
- our own router
- models
- fine-tuned in house
- compute
- gpu clusters we build
- contact
- contact@ezolabs.io
- social
- x.com/ezolabsinc
frequently asked questions
what does ezo labs do?
we build worlds for agents.then agent swarms to run them.and superbot, to manage them all.
what do you mean by a world?
everything an agent acts inside.the simulation, the tools, the rules.
what is superbot?
agents to manage your agents.in private beta at superbot.gg.
do i have to use one agent?
no. we run clusters and swarms.and mix model families in one team.
what are you building next?
it is unannounced.follow @ezolabsinc on x to hear first.
how do i reach you?
write to contact@ezolabs.io.