Amanai

Amanai Changelog

Embeddings API is live

OpenAI-compatible embeddings with BGE, EmbeddingGemma, and Qwen3 models.

New API

Embeddings API is live

Amanai now serves OpenAI-compatible text embeddings. Six launch models cover English and multilingual retrieval.

What changed

A new endpoint, POST /v1/embeddings, accepts standard OpenAI embedding requests. Existing API keys work without configuration changes — pass a bearer token or x-api-key (session cookies are not accepted for this route).

✓ Six public model IDs: BGE (small / base / large / M3), EmbeddingGemma 300M, and Qwen3 Embedding 0.6B.
✓ OpenAI-compatible request and response shape — drop-in for common embedding clients.
✓ Listed in the model catalog, dashboard calculator, documentation, and status page.

Models

Pick by dimension size and language coverage. Current rates vary by model.

ModelDimensionsLanguagesNotes
amanai/bge-small-en 384 English Fast, compact
amanai/bge-base-en 768 English Balanced
amanai/bge-large-en 1024 English Higher quality
amanai/bge-m3 1024 Multilingual 100+ languages
amanai/embeddinggemma-300m 768 Multilingual Gemma embedding
amanai/qwen3-embedding-0.6b 1024 Multilingual Qwen3 embedding

Using embeddings

Send text (string or array of strings) to /v1/embeddings:

bash 
curl https://api.amanai.dev/v1/embeddings \
  -H "Authorization: Bearer $AMANAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "amanai/bge-m3",
    "input": "hello world",
    "encoding_format": "float"
  }'

List embedding models with GET /v1/models?type=embedding (API key required). The unauthenticated GET /v1/models list stays chat and image only.

Billing

Billing has changed since this launch. BGE and Qwen3 embeddings now use model-specific per-token rates with a 250-credit minimum; EmbeddingGemma remains per UTF-8 byte with a 1,000-credit minimum. Check current embedding rates before sending a request.

Quote cost with GET /calculate?model=amanai/bge-m3&input=N, using an estimated token count for BGE/Qwen3. Use text=... to count bytes for EmbeddingGemma. The dashboard calculator also estimates embedding usage.

Failed requests before delivery are not charged. Partial delivery may settle. Historical chat and image usage is unchanged.

Availability

Embeddings appear in the dedicated Embeddings guide, Billing & pricing, credit calculator, and status page. Use any active API key — no new key is required.