Vercel AI Gateway alternative

The Vercel AI Gateway alternative hosted in Europe

The same pass-through prices with no markup on tokens, the same drop-in OpenAI compatibility, and 700+ models including image, voice and video, with the gateway itself hosted in Europe, tracing and evals built in, and no Vercel account required.

Get startedView Documentation

European end to end

The gateway runs in AWS Stockholm and routes to European providers, so the gateway leg and the inference leg both stay in Europe.

Observability included

Span-level tracing, evals and audit logs ship with the gateway, deep traces stay in one dashboard instead of leaving through metered drains.

Standalone gateway

Opper is its own account with its own API keys, no Vercel team to create and no per-request meters on tracing or governance.

Vercel AI Gateway fees and metered add-ons per vercel.com/docs/ai-gateway/pricing, model count measured from its public model endpoint, August 2026.

At a glance

Vercel AI Gateway alternatives, compared

Vercel AI Gateway vs Opper feature comparison
FeatureVercel AI GatewayOpper
Models on one unified API"Hundreds", 324 in public API700+ incl. image, voice, video
No markup on tokens
Yes
Yes
Drop-in OpenAI SDK compatibility
Yes
Yes
OpenAI, Anthropic and Gemini formats on one base URL
OpenAI and Anthropic Messages
Yes
Automatic failover across providers
Yes
Yes
Gateway hosted in Europe
Any Vercel region; pinned hosts coming
AWS Stockholm
European inference routes
Opt-in pinning, supported models
Self-serve, end to end
Prompts stored by the gatewayNone, deleted immediatelyNone, metadata only
Provider-level zero data retention
Pro/Enterprise; team-wide metered
Default on many routes, listed per route in the model directory; Enterprise covers the rest
Span-level tracing and evals built in
Traces exit via metered drains
Yes
PII and prompt-injection guards
No
Yes
Per-key and per-project spend caps
Yes
Yes
Standalone account
Requires a Vercel team
Yes
Agent CLI and AI Wallet
No
Yes

Want to compare individual models head-to-head? Run them side by side in the model playground →

Trusted by 50k+ developers and companies serving 10M+ users

Aixia
evroc
GetTested
Instabridge
LexBox
Ping Payments
Steep
Svenska Bostäder

Why teams switch

Why teams move from Vercel AI Gateway to Opper

Vercel AI Gateway is a strong default when your app already deploys on Vercel: no platform fee on tokens, prompts deleted immediately at the gateway, and zero-config auth from the AI SDK. Teams outside that platform, or with European requirements, hit four ceilings.

The gateway itself runs in any Vercel region

Regional inference, shipped in July 2026, pins the inference leg and provider-stored data to the EU, opt-in and fail-closed. The gateway leg is another matter: Vercel's own docs note a request can terminate and be processed in any Vercel region before it is forwarded to the provider, with region-pinned gateway hosts listed as coming. Gateway residency and inference residency are separate questions, and today only one of them is answered.

Governance is metered

Team-wide zero data retention and team-wide provider allowlists bill per thousand requests on Pro and Enterprise plans, custom reporting bills per write and per query, and trace drains meter both delivered events and data volume. The controls are real and well documented, but a compliance posture becomes a set of meters to budget, and Hobby plans have no zero data retention routing at all.

Deep traces live in another tool

The dashboard tracks requests by model, time to first token, token counts and spend. OpenTelemetry traces of full agent runs leave through Trace Drains into your own observability stack, metered per event and per byte on Pro and Enterprise, and evaluations are not part of the product.

Coupled to a Vercel team

Every AI Gateway account is a Vercel team account, and the plan tier decides which controls you can reach: zero data retention routing, team-wide provider allowlists and trace export all sit on Pro and Enterprise. Reaching them means holding a paid Vercel team plan even when nothing else in your stack runs on Vercel.

The Opper Way

What Opper gives you that Vercel AI Gateway doesn't

A European gateway with the production primitives included, not metered.

Hosted in Europe, end to end

Opper's gateway runs in AWS Stockholm and routes to European providers, so the request path stays in Europe from the moment it terminates to the moment a model answers. There is nothing to opt into and no plan gate, European routes are self-serve from your first call, and the gateway stores metadata only, no prompts, by default.

  • Gateway in AWS Stockholm
  • European routes end to end
  • Self-serve from the first call
Explore the LLM Gateway
Two legs to every request, both pinned
Typical gateway path
Gateway leg
any region
Inference leg
opt-in pinning
Opper path
Gateway leg
AWS Stockholm
Inference leg
European route
Gateway and inference stay in Europe
Metadata only, no prompts stored

Native tracing, evals and audit trail

Every span, every model call, every tool invocation is logged with cost, latency and user attribution. Score outputs against datasets, run evals on a schedule and export audit logs from the same dashboard that routes your traffic, with no trace drains to configure and no per-event meter to watch.

  • Span-level tracing across agent runs
  • Auto-scoring and custom evals
  • Searchable, exportable audit logs
See observability features
Live trace
span 4 of 6
classify
retrieve
generate
validate
Latency
2.4s
Tokens
1,847
Cost
$0.0092

Same OpenAI SDK, change one line

Drop-in compatible with the OpenAI client. Point the base URL from ai-gateway.vercel.sh/v1 to api.opper.ai/v3/compat, keep your prompts, tool calls and streaming code, and reach 700+ models on one key. The same endpoint also accepts Anthropic Messages and Gemini formats, and the Vercel AI SDK keeps working through its OpenAI-compatible provider support.

  • 700+ models
  • Zero rewrites
  • Automatic fallbacks
from openai import OpenAI

client = OpenAI(
  base_url="https://api.opper.ai/v3/compat",
  api_key=OPPER_API_KEY,
)

response = client.chat.completions.create(
  model="openai/gpt-4o-mini",
  messages=[...]
)

# Same SDK, 700+ models, automatic fallbacks

Guardrails at the gateway, included

PII redaction and prompt-injection detection run at the gateway before tokens reach an upstream model, alongside per-user, per-project and per-key budgets. The perimeter lives in one place rather than in application code or a set of metered add-ons.

  • PII redaction before the upstream call
  • Prompt-injection detection
  • Budgets per user, project and key
See the AI Control Plane
Gateway guards
pre-flight
Incoming prompt
Send receipt to jane@acme.com, SSN 123-45-6789
PII redacted, injection-safe
Sent to model
Send receipt to [email], SSN [ssn]
PII filters and prompt-injection detection before tokens reach upstream

Agent platform, not just a router

Opper is more than a gateway. The Agent SDK, Agent CLI and AI Wallet turn the same unified API into a platform for building, shipping and monetizing production agents, from routing a coding agent through any model to letting end users fund their own inference.

  • Agent SDK in Python and TypeScript
  • Agent CLI for Claude Code, OpenCode, Codex
  • User-funded inference via AI Wallet
Just a router
Proxy → upstream
Opper
Agent SDK

Build headless agents in Python or TypeScript

Agent CLI

Launch Claude Code, Codex, OpenCode on any model

AI Wallet

User-funded inference, no billing UI to build

Control Plane

Observe, route, steer, guard, comply

One platform, one API key

Model catalog & pricing

Access the latest models from leading providers with unified, transparent pricing per million tokens.

All prices are per 1M tokens • EU and US regions available • Prices subject to change, see docs for latest

ProviderModelRegionInput (1M tokens)Output (1M tokens)
Loading models...

Custom Models & BYOK

Bring your own API keys or add custom model deployments using the Opper CLI or API.

opper models create example/my-gpt5 azure/gpt5-production YOUR_API_KEY

Looking for a specific model? View complete model list →

FAQ

Opper vs Vercel AI Gateway FAQ

What's the difference between Opper and Vercel AI Gateway?

+
Both route hundreds of models through one key and pass provider list prices through with no markup on tokens. The differences are where the gateway runs (Opper's gateway is hosted in Europe with European routes end to end, Vercel's terminates in any Vercel region with EU pinning available for the inference leg only), what ships included (Opper builds in span-level tracing, evals, PII and prompt-injection guards, while Vercel meters custom reporting, trace drains and team-wide zero data retention as add-ons), and coupling (Opper is a standalone account, Vercel AI Gateway requires a Vercel team).

Is Opper a drop-in replacement for Vercel AI Gateway?

+
Yes. Both expose OpenAI-compatible endpoints, so you swap the base URL from ai-gateway.vercel.sh/v1 to api.opper.ai/v3/compat, change the API key, and your existing calls route through Opper. The same base URL also accepts Anthropic Messages and Gemini formats, and Opper's 700+ model catalog covers the same major model families.

Can I keep using the Vercel AI SDK with Opper?

+
Yes. The AI SDK supports custom OpenAI-compatible providers, Vercel's own gateway is simply its zero-config default. Point the SDK at Opper's compatible endpoint and keep your streaming, tool calls and framework code unchanged.

Does Vercel AI Gateway keep data in the EU?

+
Partly. Regional inference, shipped in July 2026, lets you pin the inference leg and provider-stored data to the EU, opt-in and fail-closed, for supported models and often at a regional premium. Per Vercel's own docs the gateway request itself can still terminate and be processed in any Vercel region before it is forwarded to the provider, with region-pinned gateway hosts listed as coming. Opper runs the gateway itself in Europe, in AWS Stockholm, with European routes end to end, documented at opper.ai/ai-compliance.

How does pricing compare between Opper and Vercel AI Gateway?

+
On tokens they are the same, both pass provider list prices through with no markup. Vercel charges no platform fee and meters team-wide governance features such as provider allowlists, team-wide zero data retention, custom reporting and trace drains, while Opper charges a 3% fee on credit purchases, 5.5% with Control Plane, and keeps observability and governance in the plan. The live numbers are on opper.ai/pricing.

Does Vercel AI Gateway store my prompts?

+
No. Its docs state that prompts and outputs are deleted immediately after the request completes, and a separate request flag disallows prompt training. Provider-side retention is a separate question: routing that keeps traffic on zero-retention providers is a Pro and Enterprise feature, and some models have no zero-retention provider at all. Opper stores metadata only by default and publishes the retention posture of every route in its model directory.

How do I migrate from Vercel AI Gateway to Opper?

+
Change the base URL and the API key and the first call works, model identifiers map across the catalog and nothing else has to change up front. Add European routes, guardrails or tracing when you are ready.
Let your users fund their own inference

AI Wallet

Let your users fund their own inference

The same gateway powers Opper's AI Wallet, a user-funded billing layer where end users sign in, top up a balance and pay for AI in every connected app. No metering, billing UI or per-user budget service to build.

The Vercel AI Gateway alternative hosted in Europe

Get an API key and make your first call in minutes, same models, no markup on tokens, tracing and European routes included.

Get startedView Documentation