Serverless
Pay for what you use and start sending requests immediately.
The token factory for open models
Turning open weights into production traffic.
One OpenAI-compatible API, backed by an orchestration layer built to keep open-weight models fast, available, and affordable to run at volume.
The orchestration layer between open-weight models and the traffic that runs on them.
Cost-efficient by design
Serving efficiency is the product.
Day-zero availability
Ready to serve the moment a strong release lands.
One fabric, many providers
Resilience without the integration burden.
Platform
Serve open models your way
From a first request to reserved capacity, the platform scales with how you build.
Dedicated
Provisioned capacity for production workloads.
Enterprise
Guaranteed capacity and custom commitments.
Model catalog
A broad, current catalog of open weights
Chat, code, vision, and embedding models across the fastest-moving open-weight families.
Day-zero
New open-weight releases go live the moment they're ready to serve.
Kimi
Long-context chat models for document-heavy workloads.
Chat
GLM
Bilingual chat and reasoning models.
Chat
Reasoning
Llama
General-purpose chat and instruction-following models.
Chat
Code
DeepSeek
Reasoning-forward models tuned for complex problem solving.
Chat
Reasoning
Qwen
Multilingual chat and code models.
Chat
Code
Vision
Embeddings
Mistral
Efficient, general-purpose models built for high-throughput serving.
Chat
Code
Solutions
Built for what comes after serving
Domain tuning, vertical-specific models, and the infrastructure agents run on.
Post-trained models
Models post-trained on your domain and data.
Industry-specific models
Model variants adapted for concrete verticals.
Agent infrastructure
Frameworks, evaluations, and workflow orchestration for agents.
AI Gateway
Generally availableThe capability you asked for most, now generally available
One governed surface for every model call.
Cost controls
Budgets, limits, and attribution on every model, harness, or agent.
Guardrails, auditing & ACLs
Guardrails, audit trails, and access-control lists across all AI traffic.
Every model, one gateway
Closed and open models, behind the same governed surface.
Infrastructure
One fabric. Many providers.
A multi-provider, multi-region compute fabric, presented as a single, simple API.
Multi-provider, multi-region
Capacity is drawn from many providers and regions and presented as one surface.
Latency and reliability, engineered
Fast time to first token and steady generation are first-class commitments.
Hardware diversity by design
Workloads are matched to the hardware they run best on.
Developers
Drop-in OpenAI-compatible API
Point your existing client at the platform and start serving open models.
- Drop-in compatibleWorks with existing OpenAI-compatible clients and SDKs out of the box.
- Streaming built inToken streaming for responsive, incremental output in production.
- Simple onboardingGet an API key and start sending requests — no infrastructure to provision.
curl https://api.example.com/v1/chat/completions \
-H "Authorization: Bearer $API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "your-model-id",
"messages": [{"role": "user", "content": "Hello"}],
"stream": true
}'from openai import OpenAI
client = OpenAI(
base_url="https://api.example.com/v1",
api_key="YOUR_API_KEY",
)
response = client.chat.completions.create(
model="your-model-id",
messages=[{"role": "user", "content": "Hello"}],
stream=True,
)
for chunk in response:
print(chunk.choices[0].delta.content or "", end="")Start building on open models today
Bring your traffic. We handle the fabric underneath.