Groups

PulseNeko5/20/26About 1 min

Every API key binds a group. The group decides:

Which models this key can reach
Which upstream channel routes the request
The billing multiplier (group ratio)

Current groups

Group key	Name	Positioning
`default`	Default	Generic fallback
`anthropic`	Anthropic Official	Claude family via Anthropic direct
`openai`	OpenAI Official	GPT family via OpenAI direct
`google`	Google Official	Gemini family via Google direct
`cc`	Claude Code	Re-uses Claude Code subscription, cheaper
`aws-q`	AWS Q	Re-uses AWS Q, discounts on Claude / Opus 1M
`codex`	Codex	Re-uses Codex subscription, GPT-5 codex discount
`windsurf`	Windsurf	Re-uses Windsurf channel

Real multipliers & models

Multipliers shift with promos / free-tier phases. The truth is the multiplier shown next to each option in the Create API key dialog.
Per-group model availability lives on Model Catalog — each model card carries a group tag.

Picking a group

By the model you use most:

Heavy Claude Code user → check cc / aws-q first
Heavy Codex CLI user → codex
All-in OpenAI ecosystem → default / openai
Gemini-heavy → default / google
Fall back to default / the official group if a discount group doesn't carry your model

Mixed workloads → multiple keys, one per group is the clean pattern.

Switching

Existing keys can't change group (anti-misbinding). Create a new key, delete the old one.

Model not in this group?

You get:

{
  "error": {
    "type": "invalid_request_error",
    "message": "model 'xxx' is not available for your token group"
  }
}

Look up the model on Model Catalog, check its Group tag, create a key in that group.

"Deprecated" groups

The announcement bar may say "group X is free / paused / changed". This affects current pricing / availability for that group — read the announcement for specifics.