New
Halo 2 is out
Read the release
Models that reason,
build, and imagine.
One family of models for language, code, speech and images — trained and served on compute we design ourselves.
Why do cats knead blankets?
It’s a kitten habit that stuck — kneading helped them nurse, and now it signals they feel safe.
Explain compound interest in one line
You earn interest on your interest, so growth speeds up the longer you wait.
Name ideas for a small bakery?
Rise & Rye, Crumb Theory, or Morning Loaf — short, warm, easy to remember.
How far away is Mars right now?
Somewhere between 55 and 400 million km, depending on where both planets sit in orbit.
Make this email sound friendlier
Sure — I softened the opening and swapped “per my last note” for “just circling back.”
What’s the tallest tree on Earth?
A coast redwood nicknamed Hyperion, roughly 116 meters tall.
Assist
Explore
forge — ~/studio-app
› Add rate limits to the upload route
planning…
open api/upload.ts
search “quota” · 3 hits
edit api/upload.ts +18 −2
run npm test
✓ 24 passing · 1.9s
› Move image resizing to a queue
thinking… 3.1s
open workers/resize.ts
create queue/jobs.ts
edit api/images.ts +42 −17
agent review changes · done
✓ build passed
› Write docs for the new endpoint
open docs/api.md
edit docs/api.md +64
✓ ready for review
Forge
Explore
My order showed up cracked. Can I get a refund?
Relay is working
Relay
Explore
Studio
Explore
Speak
Explore
For developers
Build once.
Ship everywhere.
A single API for text, code, speech and images. Swap models with one parameter and scale without rewriting a line.
Create a key
API reference
3B+
tokens served daily
180ms
median first token
6
model families
Copy
from vantir import Vantir
client = Vantir()
reply = client.responses.create(
model=“halo-2”,
input=“Draft a launch checklist for our iOS app”,
)
# streamed, typed, and billed per token
print(reply.output_text)
Python
TypeScript
Go
cURL
12B
requests answered last month
250K
accelerators in our fleet
Latest news
Halo 2
Product
·
Introducing Halo 2
Relay in the clinic
Customers
·
How a clinic network cut triage time in half with Relay
Speak 3
Product
·
Speak 3: faster, warmer, and fluent in 40 languages
Memory in Forge
Engineering
·
Long-term memory comes to Forge
Pick your starting point
Build it yourself
Everything you need to ship a product on Vantir today.
Every Vantir model, one key
Pay only for the tokens you use
Rate limits that grow with you
Guides, SDKs and sample apps
Start building
Work with our team
Dedicated help and custom terms for larger rollouts.
Guided onboarding with an engineer
Custom throughput and limits
Monthly invoicing
SSO, audit logs and data residency
Talk to sales