Run your existing OpenAI-compatible tools on Verdict's models
Point the OpenAI SDK, curl, or your agent CLI at one new base_url and keep working the way you already do.
from openai import OpenAI
client = OpenAI(base_url="https://api.verdictide.com/v1", api_key="vrd_live_...")
client.chat.completions.create(model="verdict/verdict-4-20b", messages=[{"role": "user", "content": "..."}])Direct API access is in private beta.
Model & pricing
verdict-4-20b
Input: $0.03 / MTok
Output: $4.00 / MTok
Billed in credits from the same wallet as Verdict IDE and Verdict Chat.
Prices are served live from GET /v1/models, so this page stays accurate as models are added.
What you get
Streaming, OpenAI-shaped
Server-sent events on every completion, so an existing streaming client keeps working unmodified.
Usage on every stream
The final chunk carries the usage block, so token accounting doesn't require a second request.
Rate-limit headers
x-ratelimit-* headers on every response tell a client exactly how much headroom is left before it has to ask.
Keys that expire on their own
Every key can be set to expire automatically -- 30, 90, or 365 days, or never -- so a stale key retires itself instead of sitting live indefinitely.
Compatibility, made explicit
Verdict's Direct API is OpenAI-compatible per the published compatibility matrix: every request field is documented as supported, ignored, or rejected -- nothing is silently dropped.
View the compatibility matrix