DEPLOY AI
ABOVE THE
CLOUDS
SHIP MODELS TO 40 REGIONS IN ONE COMMAND
— NO MATTER WHERE YOUR USERS ARE
TRUSTED BY TEAMS SHIPPING AI AT SCALE
NORTHWIND
HALCYON
VERTEX/LABS
OKTAVE
MERIDIAN
LUMEN.AI
NORTHWIND
HALCYON
VERTEX/LABS
OKTAVE
MERIDIAN
LUMEN.AI
BUILT FOR THE
HIGHEST LOADS
ONE COMMAND DEPLOY
Point Altiq at a model and a region list. It builds, warms GPUs and hands you an endpoint.
40 REGIONS, ONE ENDPOINT
Requests land on the closest healthy GPU. No load balancer to write.
SCALE TO ZERO
Idle models cost nothing. Cold starts stay under two seconds.
OBSERVABILITY BUILT IN
Tokens per second, p95 latency and every request log in one view.
FROM LAPTOP
TO 40 REGIONS
TEAMS THAT
SHIP FASTER
“We moved our inference off three clouds in a weekend. Latency dropped by half.”
Jonas W.
CTO, NORTHWIND
“Scale to zero paid for itself in the first month. Nights and weekends are free now.”
Leila M.
ML LEAD, HALCYON
“The terminal is the product. One command and we are live in 40 regions.”
Kenji O.
FOUNDER, OKTAVE
PAY FOR TOKENS,
NOT IDLE GPUS
HOBBY
$0
/ month
$10 in free credits every month.
Scale to zero
3 regions
Community support
PRO
POPULAR
$49
/ month + usage
For teams running models in production.
All 40 regions
Autoscale to 512 GPUs
Metrics + request logs
Private networking
Email support, 4h
SCALE
Custom
Reserved capacity and a direct line to us.
Dedicated GPUs
99.99% SLA
SOC 2 + HIPAA
Solutions engineer
QUESTIONS,
ANSWERED
Which models can I deploy?
[ – ]
Anything with open weights: Llama, Mistral, Qwen, Whisper, SDXL, or your own fine-tune in a container.
How fast are cold starts?
[ + ]
Under two seconds for most models. Weights stay cached on local NVMe, so an endpoint scaled to zero wakes up before your user notices.
Do you train models too?
[ + ]
Not yet. Altiq is built for inference. Bring weights from any training stack and we serve them in every region you list.
Where is my data stored?
[ + ]
Only in the regions you pick. Requests and logs stay inside that region, encrypted at rest and in transit.
ALTIQ EDGE
Run models on GPUs 20 ms from your users, in 300 cities.
Join the waitlist for the private beta.
1,920 TEAMS ON THE LIST