1536 / 3072 dimensions
Two tiers. Small for retrieval at scale, large for nuanced semantic search.
Sub-millisecond P50
Batched inference and a custom kernel keep latency low at 10K vectors per second.
OpenAI-compatible
Same /v1/embeddings endpoint. Same request shape. Switch by changing the base URL.
Multi-modal
Text, image, audio: same vector space. Cross-modal search out of the box.
100+ languages
Multilingual by design. No separate model per language.
Batched privacy
Embeddings are not stored unless you opt in. We compute, return, and forget.
Pricing
Token-efficient, with volume discounts and burn incentives.
Best for retrieval at scale
Best for nuanced semantics
Text + image + audio
Async batch API
Pricing is illustrative. Final rates are governed by on-chain parameters and may vary based on network state.
Staking requirements
Tier-based access. Higher stakes unlock better economics and more capacity.
| Tier | Required stake | Access |
|---|---|---|
| Explorer | 10M QUBIC | Best-effort, 60 RPS |
| Builder | 50M QUBIC | Reserved, 600 RPS |
| Startup | 150M QUBIC | Dedicated, 6K RPS |
| Business | 500M QUBIC | Burst to 60K RPS |
| Enterprise | Custom | Unlimited |
Example
Drop-in compatible with the OpenAI SDK.
from aigarth import Aigarth
client = Aigarth(api_key="sk-aigarth-...")
texts = [
"What is Useful Proof of Staking?",
"How does Aigarth work?",
"Best pizza in New York",
]
response = client.embeddings.create(
model="aigarth-embed-1-large",
input=texts,
)
vectors = [item.embedding for item in response.data]
print(f"Embedded {len(vectors)} texts at {len(vectors[0])} dimensions")Enterprise benefits
Everything in the standard tier, plus the things enterprises need.
- Dedicated vector index capacity
- Co-located with managed Pinecone, Weaviate, Qdrant
- Migration from OpenAI, Cohere, Voyage in days
- HIPAA-aligned data handling
- PII redaction in pre-processing pipeline
- Multi-region replication for global apps
- SOC 2 and ISO 27001 (in progress)
- 24/7 on-call for production incidents