One API for every model.
Pay what the model costs.

Get a key Read the docs Compute prices

curl https://api.openvibe.exchange/v1/chat/completions \
  -H "Authorization: Bearer $OPENVIBE_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"openvibe/auto","messages":[{"role":"user","content":"Hello"}]}'

models: 9
tokens routed this week: 16,882,288
median reply: 1,974 ms

Popular models

GPT-5 nanoGPT
In / out per M
$0.05 / $0.4
Uptime
55.19%
Speed
1,699 ms
Try in Chat How it ranks
GPT-5 miniGPT
In / out per M
$0.25 / $2
Uptime
53.38%
Speed
4,209 ms
Try in Chat How it ranks
DeepSeek V4.1 FlashDeepSeekFree to try
In / out per M
$0.3 / $1.2
Uptime
44.9%
Speed
12,974 ms
Try in Chat How it ranks
DeepSeek V4 ProDeepSeek
In / out per M
$1.32 / $3.96
Uptime
–
Speed
–
Try in Chat How it ranks
GPT-4.1 miniGPT
In / out per M
$0.4 / $1.6
Uptime
–
Speed
–
Try in Chat How it ranks
GPT-4o miniGPT
In / out per M
$0.15 / $0.6
Uptime
–
Speed
–
Try in Chat How it ranks
GPT-5GPT
In / out per M
$1.25 / $10
Uptime
–
Speed
–
Try in Chat How it ranks
Muse Ember 1.0muse-ember
In / out per M
$0.1 / $0.5
Uptime
–
Speed
–
Try in Chat How it ranks
Muse Spark 1.3muse-spark
In / out per M
$1.25 / $4.25
Uptime
–
Speed
–
Try in Chat How it ranks

New this week

Every model spotted

0% fee on your own key

Bring your provider key and pay the provider's price, nothing more. No cap, no end date.

Add a key

The cheapest route

Every call goes to the cheapest provider that is up. Open models on OpenVibe's own GPUs, at a fixed price per million tokens, are coming next.

Compute prices

Uncensored, 18+

Uncensored models for adults who opt in. The content floor applies on every request.

What the floor covers

Pricing

Your own provider keyThe provider's price. OpenVibe adds 0%, with no cap and no end date.
Managed modelsThe input and output price per million tokens shown on each model, before you call.
Open models on OpenVibe's GPUsA fixed price per million tokens, the same on every call. Coming next.
Your own endpointAn hourly price, quoted before it starts; it stops when idle.
SearchEach engine's price per 1,000 searches; the free engines first.

Paid models open soon. Every price shows before you call, and nothing is charged until then.

Questions

Can I use my existing client?

Yes. Point any OpenAI-compatible client at the Exchange endpoint with your Exchange key; Anthropic-style clients work through /v1/messages.

Read the docs

Where can I talk to a model?

In Vibe on OpenVibe.Chat: talk to any model here, or two side by side.

Open Vibe

Where do the numbers come from?

Prices come from the live catalogue. Speed and reliability come from measured requests. A number nobody has measured stays unknown, never guessed.

See the status

One API for every model.Pay what the model costs.