Open models, one API, no GPUs to babysit
Openweight serves open-weight language, embedding and image models behind an OpenAI-compatible endpoint. Pay per token, scale to zero, bring your own fine-tunes.
Small, medium or large — same endpoint
Swap the model name, keep the code.
Fast and cheap for classification, extraction and routing.
Change the base URL and you're done
Pick a language, copy the snippet, paste it in.
from openai import OpenAI
client = OpenAI(base_url="https://api.openweight.dev/v1", api_key=OW_KEY)
reply = client.chat.completions.create(
model="general-70b",
messages=[{"role": "user", "content": "Summarise this ticket in one line"}],
)Where each request spends its time
Relative time per stage for a streamed chat request.
Platform
Everything around the model
Scale to zero
Pay nothing when traffic stops; cold starts are kept short.
Bring your fine-tune
Upload LoRA adapters and serve them on shared base models.
Regional endpoints
Keep requests and logs in the region you choose.
No training on traffic
Prompts and outputs are never used to train anything.
Usage dashboards
Tokens, latency and errors per key and per model.
Rate limits you set
Per-key limits so one service can't starve the rest.
Pricing
Per million tokens, no minimums
Free tier
Free
Pay as you go
$0.6/unit
Growth
$0.45/unit
Committed
$0.3/unit
FAQ
Developer questions
Filter by what you're trying to figure out — billing, security, or how the product fits your stack — then expand a question for the full answer.
One key, every open model
Create a key and send your first request.
© 2026 Openweight