Anthropic
Claude — strong reasoning and long-context coding.
Docs · Core concepts
Choosing a model.
ubc now proxies to the major frontier providers behind one OpenAI-compatible endpoint. One daily budget covers every one of them.
Providers
Eight families, one budget.
ubc now routes to models across these provider families. Availability tracks what the upstream serves.
OpenAI
GPT family — broad general-purpose capability.
xAI
Grok — fast, current, conversational.
Gemini / Gemma — multimodal and efficient.
DeepSeek
Strong open-weight reasoning and code.
Kimi / Qwen
Moonshot and Alibaba long-context open models.
Zhipu GLM
The GLM family from Zhipu AI (Z.ai) — strong reasoning and code.
Selecting
How to pick a model.
In browser chat
Open the model picker, choose a provider and variant, then chat. Switch any time; the budget is shared.
Over the API
Set the model field to an exact model id. The id must match one ubc now serves — it does not alias names like "gpt-4o".
Discover the ids
List the catalog from the OpenAI-compatible /models route and copy an id verbatim.
List the models ubc now serves
curl https://ubcnow.ai/v1/models \-H "Authorization: Bearer ubc_sk_…"
Good to know
Questions about models.
Which model should I use?
For coding, a strong Anthropic or DeepSeek model is a good default. For quick general chat, any GPT or Grok variant works. They all draw from the same budget, so experiment freely.
How are tokens counted?
Text usage is metered by the model's token count for the request and response, deducted from your daily budget. Image generation is a flat 5,000 tokens per image.
What happens when I run out?
Requests return a 402 until the budget resets at 00:00 UTC. You can earn extra credits via quests on the usage page.
Try every frontier model.
One verification, one budget, the whole field.