Platform API
An OpenAI-compatible API for cognira Retrain.
Pay as you go for cognira Retrain 12B and Omega 188M at £1.50 per million tokens. Set base_url to https://cognira.dev/v1. The request field is still conway-retrain — that is the id the API expects. Keys and billing live in the signed-in dashboard.
Models
conway-retrain
cognira Retrain 12B, served through MLX. Reasoning, when present, comes back as
reasoning_content.conway-omega
Omega 188M, the earlier compact checkpoint, billed at the same rate.
Quickstart
from openai import OpenAI
client = OpenAI(
api_key="conw_live_...",
base_url="https://cognira.dev/v1",
)
response = client.chat.completions.create(
model="conway-retrain",
messages=[{"role": "user", "content": "hello"}],
)
print(response.choices[0].message.content)POST /v1/generate accepts { model, message, max_tokens }. Pass stream: true on chat completions for Server-Sent Events.
Questions
- Is the cognira API OpenAI-compatible?
- Yes. Point any OpenAI SDK at https://cognira.dev/v1 and send a cognira API key. The model id for cognira Retrain 12B is conway-retrain. Omega 188M is conway-omega.
- How much does the Platform API cost?
- £1.50 per million tokens, input and output, for either model. The minimum top-up is £5. Chat plans (Free, Pro, Max) are a separate monthly subscription.
- Which models can I call?
- conway-retrain is cognira Retrain 12B, the model behind the chat. It can return reasoning_content. conway-omega is Omega 188M, still served at the same price, with no chain of thought.
- Does API traffic train the model?
- A reply only enters shared training after you approve it with POST /v1/feedback and a rating of up. Silence or a thumbs-down does not.