Northvane

● ROUTING 2.4B TOKENS / DAY

Every prompt,
on the fastest GPU.

One OpenAI-compatible endpoint. Northvane scores every provider on live latency, price and health, then sends each request down the fastest route.

Start routingRead the docs

$pip install northvane

  • Northvane
  • Fieldnote
  • Halcyon
  • Quillo
  • Orbital
  • Kestrel
  • Tessel
  • Veyra
  • Lumora
  • Driftwave
  • Nimbra
  • Ledgerly
  • Corvid Labs
  • Brisa
  • Northpeak
  • Oakline
  • Arcwell
  • Monoline

ROUTING TRAFFIC FOR TEAMS SHIPPING AI IN PRODUCTION

A routing layer for the GPU era. Providers change prices, slow down at peak hours and go offline without warning. Northvane absorbs every one of those moments, so your product never has to.

Latency spikes3×

Peak hours can triple response times on a single cloud.

Price changes+40%

Your margin moves every time their pricing does.

Outages4 h

When one provider goes down, your product goes with it.

01 · BUILT FOR SPEED

Faster than any single provider.

  • Scores six providers in under a millisecond
  • Learns each region’s peak hours
  • Falls back before a timeout ever fires

P50 LATENCY · LLAMA-3.1-70B · LAST 5 MIN

Cloud B

96 ms

Cloud A

84 ms

Cloud E

79 ms

Northvane (routed)

71 ms

02 · LIVE PROVIDER BOARD

The fastest route changes
every second.

Northvane re-ranks every provider continuously. The route you get is the best one right now, not the best one last quarter.

#PROVIDERREGIONP50 LATENCY$ / 1M TOKUPTIME 30DLOADTREND
01Cloud CACTIVE ROUTEus-west71 ms$0.4299.99%██████░░░░▃▂▂▁▂▁▁▂
02Cloud Eus-central79 ms$0.5199.97%███████░░░▂▃▂▃▂▂▃▂
03Cloud Aus-east84 ms$0.4899.95%████████░░▃▄▃▂▄▃▄▃
04Cloud Beu-west96 ms$0.3999.90%█████░░░░░▄▃▅▄▃▄▅▄
05Cloud Dap-south112 ms$0.3599.80%████░░░░░░▅▆▄▅▇▅▆▅
06Cloud Feu-north138 ms$0.3199.70%███░░░░░░░▆▅▇▆▅▇▆█

POLICY · LOWEST LATENCY UNDER $0.50 / 1M

LIVE · P50 OVER LAST 5 MIN · REFRESHED 1 S AGO

FIELD NOTES

“We cut our inference bill by a third in the first week, and stopped getting paged at 3 a.m.”
MOMara OkaforHEAD OF PLATFORM · FLUIDNOTE

03 · DROP-IN

Two lines.
Nothing else changes.

Keep the OpenAI SDK you already use. Point it at Northvane and set the provider to auto; the router picks the winner on every call.

2
LINES CHANGED
0
NEW SDKS
<5
MINUTES TO SHIP

route.py

01from openai import OpenAI02 03client = OpenAI(04 base_url="https://api.northvane.dev/v1",05 api_key=os.environ["NORTHVANE_KEY"],06)07 08client.chat.completions.create(09 model="llama-3.1-70b",10 extra_body={"provider":"cloud-a""auto""cloud-c"},11)

04 · CAPABILITIES

Quietly handling the hard parts. Observability, failover, caching and budgets, built into the route.

Live observability. Every request traced with its provider, latency, cost and fallback path.

P50 · 71 MS

Automatic failover. A degraded provider is skipped before your users notice.

CLOUD A → CLOUD C

Smart caching. Repeated prompts are served from cache at zero provider cost.

84% HIT RATE

Budget caps. Set a monthly ceiling; routing gets cheaper as you approach it.

$4.2K / $6K

Region pinning. Keep traffic inside the data boundary you choose.

EU-WEST · PINNED

05 · PRICING

A small fee on every token.
A better bottom line on all of them.

HOBBY

Free

Up to 1M routed tokens a month

  • 3 providers
  • Latency routing
  • Community support
Start free

SCALE

MOST POPULAR

$0.08

per 1M routed tokens

  • Unlimited providers
  • Failover and caching
  • Budget caps
  • Email support
Start routing

ENTERPRISE

Custom

Volume pricing and SLAs

  • Region pinning
  • SSO and audit logs
  • Dedicated routing
  • 99.99% SLA
Talk to us

Point your traffic north.

FREE UP TO 1M TOKENS · NO CREDIT CARD