API models and modes
The curated model list, how to choose a model, the anonymous and TEE modes, and the pricing fields returned by /v1/models.
Ask GET /v1/models for all models open to your key. The reply follows the OpenAI format and adds Angerona's own fields:
curl https://api.angerona.chat/v1/models -H "Authorization: Bearer $ANGERONA_API_KEY"
{
"object": "list",
"data": [
{
"id": "claude-sonnet-5.5",
"object": "model",
"owned_by": "network",
"name": "Claude Sonnet 5.5",
"mode": "anonymous",
"type": "text",
"contextLength": 1000000,
"pricingKind": "market",
"listPrice": { "metering": "per_token", "inputPer1MUsd": 2, "outputPer1MUsd": 10 },
"marketPrice": { "metering": "per_token", "inputPer1MUsd": 1.6, "outputPer1MUsd": 8, "updatedAt": "2026-10-27T19:00:00.000Z" },
"capabilities": { "streaming": true, "vision": true, "tools": true, "reasoning": true },
"presets": ["Smart"],
"popularRank": 16
}
]
}
The catalog is short on purpose
Hundreds of options would help nobody, so the list stays tight:
- Chat models: the families people use most, with one coding model and one uncensored model added.
- Image and video models: a small set for each.
popularRank orders models inside each type by recent usage, with 1 at the top. presets flags sensible defaults:
| Preset | Text | Image |
|---|---|---|
| Fast | deepseek-v4.1-flash | flux.2-klein-4b |
| Smart | claude-sonnet-5.5 | seedream-4.5 |
| Top | claude-opus-5.5 | gemini-2.5-flash-image |
| Uncensored | deepseek-v4.1-flash-uncensored | None |
| Vision | gpt-6-luna | None |
Read /v1/models at runtime instead of hard-coding ids.
API keys run in Anonymous mode
| Anonymous | TEE | |
|---|---|---|
| Through a key | Yes (default) | No |
| Hides | Your identity | Your identity and your words (end-to-end encrypted, attested hardware) |
| Price | Text is pricingKind: "market": never above listPrice, usually below. Images and video are exactly listPrice | pricingKind: "list": exactly listPrice |
Anonymous applies by default, so mode is optional. A secondary compute market of trusted providers serves the call, picking the best available trusted offer for your model and falling back if that offer fails before replying. The provider reads your prompt with no idea who wrote it, so leave out what a provider shouldn't see.
TEE puts open models in hardware enclaves, closing your request even to us. You'll find it in the Angerona app. Through a key, "mode": "tee" returns 403 tee_api_not_available. @angerona/sdk ships the complete encrypted flow.
Familiar OpenAI parameters work
| Parameter | Notes |
|---|---|
messages | Required |
stream | Supported |
max_tokens / max_completion_tokens | The default is 4096. Reasoning models get at least 1024 |
temperature, top_p, stop, seed, presence_penalty, frequency_penalty | Passed through |
response_format, reasoning_effort | Passed through |
tools, tool_choice | Passed through |
Two extra fields are Angerona's own:
| Field | What it does |
|---|---|
fallback_models | At most 2 backup ids, tried in order when your model can't reply. The x-angerona-model response header reveals which one served you |
mode | "anonymous" (default). "tee" runs in the app only |