New Release
Unify 50+ AI providers behind one API. Route intelligently by latency, cost, and availability. Ship production workloads with confidence.
831K+
Monthly tokens
100+
Models
15+
Providers
Trusted by teams using
Standardise providers behind a single API, then enforce global policies across latency, cost, capability, and compliance requirements.
Keep one integration while you swap providers, add new models, and expand modalities without rewriting clients.
Health-aware routing, automatic failover, and guardrails that protect production traffic when a provider degrades.
Route by latency, price, geography, capability, or your own rules. Apply consistent behaviour across every provider.
Model retirement and breaking changes become predictable. Gateway surfaces alerts and migration paths before downtime.
Text, vision, audio, video, embeddings, tool-calling, and realtime are treated as first-class, not bolt-ons.
Scoped keys, audit trails, and encrypted BYOK workflows. Built for teams who can't afford surprises.
Keep your existing SDK ergonomics, or adopt AI Stats SDKs for typed routing and first-class observability.
Keep your SDK, move traffic to Gateway, and standardize routing and policies.
Vercel AI SDK
Provider + model routing through AI SDK adapters.
OpenAI SDK
Drop-in base URL swap with existing OpenAI code.
Anthropic SDK
Native Anthropic SDK support with compatibility shims.
Works with
Base URL + API key migration path, with typed routing controls available when you need deeper policy logic.
Keep native SDK calls, route through AI Stats Gateway.
OpenAI SDK
Drop-in replacement: change base URL and keep existing request shapes.
import OpenAI from "openai";
const client = new OpenAI({
apiKey: process.env.AI_STATS_API_KEY!,
baseURL: "https://gateway.ai-stats.phaseo.app/v1"
});
const response = await client.chat.completions.create({
model: "openai/gpt-5.2",
messages: [
{ role: "user", content: "Summarize the deployment status." }
]
});Requests keep native SDK semantics while Gateway applies routing policy, telemetry, and fallback logic.
Optimised routing and observability built in - no homegrown adapters, no hidden markups.
| Capability | AI Stats Gateway | OpenRouter | Vercel AI SDK |
|---|---|---|---|
Model Coverage Number of models and providers supported | 500+ models, 50+ providers Largest verified catalogue updated nightly | 200+ models Varies by provider availability | Bring your own Manual adapter setup required |
Modalities Supported input and output types | Text, Vision, Audio, Video, Embeddings First-class multimodal support | Text, Vision Limited modality support | Text, Vision Provider-dependent |
Routing Intelligence How requests are distributed across providers | Latency, cost, error-aware Deterministic fallbacks with circuit breakers | Priority ordering Manual fallback configuration | Basic Limited multi-provider routing |
Observability Built-in monitoring and analytics | Full-stack telemetry Live dashboards, alerts, cost tracking | Basic analytics Requests and spend only | Self-managed Requires external tools |
Pricing Model Credit purchase fee structure | 7% basic, 5% enterprise top-up fee Applied on credit purchases, not token requests | 5.5% flat Fixed rate for all usage | 0% platform fee But limited routing capabilities |
Everything you need to know about the AI Stats Gateway. Cannot find an answer? Reach out to our team.
Yes - bring your own keys (BYOK) for any provider and enforce limits per key. Your keys are encrypted at rest and never logged. Managed keys are also available if you prefer not to handle credentials yourself.
We support chat completions, embeddings, moderations, image generation, audio generation, and video generation - each with consistent schemas across all providers. All modalities are first-class citizens with full routing and observability support.
New provider models and community submissions are reviewed and added on a rolling basis - typically within 24-48 hours of release. Once approved, they are available instantly in the Gateway and the SDKs with no code changes required.
All telemetry comes from live Gateway traffic. The dashboards display real-world token usage, latency percentiles, and provider health scores. This data feeds directly into our intelligent routing algorithms.
Absolutely - choose the provider or model per request while keeping a single unified API. Switching providers requires only changing the model name. You can even set up automatic fallbacks between providers.
Routing is evaluated on a request-by-request basis using real-time provider health, latency metrics, and your configured policies. You can route by lowest latency, lowest cost, specific regions, or custom rules. Circuit breakers automatically redirect traffic away from degraded providers.
No - we do not impose rate limits. Upstream providers may have their own limits, but we work with providers to optimize your allocation. You can optionally set spend limits per API key to control costs.
We follow industry best practices for data protection. All data is encrypted in transit and at rest. We support SSO, audit logging, and fine-grained access controls for enterprise deployments.
Unify provider access, enforce intelligent routing policies, and keep SLAs stable as the model landscape shifts. Start with our free models - scale when you are ready.
Make your first API call in less than 5 minutes with our quickstart.