Compare gateways

Experiential vs OpenRouter

Both are hosted gateways with one key, and Experiential keeps OpenRouter's paths and ids while adding native Gemini, per-response ZDR verdicts and no fee on your own keys.

このページはまだ翻訳されていません。英語版を表示しています。

Summary

OpenRouter is a hosted marketplace with 600+ models and 80+ providers. It charges a fee when you buy credits, and on your own keys above a monthly allowance. Experiential serves the same OpenRouter paths and model id spellings, so most projects move by changing the base URL and key.

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://openrouter.ai/api/v1",
    api_key=os.environ["OPENROUTER_API_KEY"],
)
client.chat.completions.create(
    model="openai/gpt-6-luna",
    messages=[{"role": "user", "content": "Hello"}],
    extra_body={"usage": {"include": True}},
)

Side by side

ExperientialOpenRouter
APIsChat, Responses, Messages, Gemini generateContent, embeddings, batchesChat, Responses, Messages, embeddings, batches
Native Gemini APIYes, every modelNot documented
Price on tokensList price, no markup5.5% fee on credit purchases (Standard), 8% (Business)
Bring your own keyNo markup; first paid key needs ProNo fee on the first $25K a month, then 5%
Models980 listed, all callable600+
FailoverOrdered waterfall; slow lanes demoted and recovered on their ownProvider fallbacks, models fallback list
Zero data retentionPro: org switch or provider.zdr; x-gateway-zdr on every responseOrg setting or provider.zdr
No-training routingEvery planPrivacy setting
Spend limitsPer-key daily, weekly, monthly, yearly caps; budgetsPer-key limits, guardrail budgets
Content guardrailsStandard pack or your own classifier, per keyBudget, access and moderation policies
SSONot yet availableEnterprise

Where Experiential is better

  • No fee on your own keys, at any volume. See Plans and pricing.
  • Proof on every response. x-gateway-provider and x-gateway-zdr name the provider that served the request and its zero-retention verdict. See Data controls.
  • Native Gemini at /v1beta beside native Messages, reaching every catalog model.
  • Slow lanes are demoted automatically, not just failed ones. See Routing and failover.

When OpenRouter may fit better

You need SSO or SAML today, or a provider we do not list.

Migration notes

WhatOpenRouterExperiential
Base URLhttps://openrouter.ai/api/v1https://api.experientiallabs.ai/api/v1 (same paths) or /v1
Anthropic SDK basehttps://openrouter.ai/apihttps://api.experientiallabs.ai (no /v1)
AuthAuthorization: Bearer sk-or-v1-...Authorization: Bearer YOUR_API_KEY
Model idsvendor/modelvendor/slug works when listed in GET /api/v1/models; use its canonical_slug otherwise
Variants:nitro, :floor, :online, :thinkingRemove them; only :free and, where a model lists it, :fast exist
usage: {"include": true}Adds usageAccepted and ignored; usage and usage.cost are always there
provider objectRouting preferencesIgnored, except provider: {"zdr": true}
models fallback listModel fallbacksUse the first model; failover is automatic
HTTP-Referer, X-TitleApp attributionAccepted, not shown
/generation, /credits, /auth/key, /activityAccount reads/generation, /credits, /key, /activity under /api/v1

Switch with an agent

Paste this into your coding agent. It inventories the project first and asks before anything it cannot map.

GET/api/v1/models
curl https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "openai/gpt-6-luna:nitro",
       "messages": [{"role": "user", "content": "Hello"}],
       "provider": {"order": ["azure"], "zdr": true},
       "usage": {"include": true}}'

Sources, checked 2026-11-19

  1. OpenRouter pricing: plans, 5.5% and 8% platform fees, BYOK allowance, 600+ models, 80+ providers, SSO on Enterprise.
  2. OpenRouter FAQ: credit fee, BYOK fee, data retention tiers.
  3. Anthropic Messages, Responses, embeddings, batch API.
  4. Provider routing, ZDR, guardrails, prompt caching.
  5. Experiential: /llms.txt, GET /api/models (total: 980).