Last updated: October 20, 2026 · Facts checked against OpenRouter's pricing page and FAQ, LiteLLM's pricing page, docs and GitHub, and Oxyy's live pages on this date.
Disclosure: This comparison is published by the Oxyy.ai team. OpenRouter and LiteLLM facts come from their own public pages.
Quick answer: OpenRouter is a managed gateway that bills each provider's list price plus a 5.5% fee on credit purchases. LiteLLM is a free, open-source proxy you host yourself: no platform fee, but you pay list price to every provider plus your own servers and maintenance. Oxyy.ai is a managed gateway that currently charges 40% of list price, so $1,000 of usage costs about $1,055 on OpenRouter, $1,000 plus infrastructure on LiteLLM, and $400 on Oxyy.
| Factor | OpenRouter | LiteLLM (self-hosted) | Oxyy.ai |
|---|---|---|---|
| Type | Managed gateway | Open-source proxy you run | Managed gateway |
| Token price | Provider list price | Provider list price (your own keys) | 40% of list price (current promo) |
| Platform fee | 5.5% on credits (Standard), 8% (Business) | $0 (open source); Enterprise license sized by request capacity | None |
| You pay on $1,000 of list-price usage | ~$1,055 | $1,000 + servers + ops time | $400 |
| Models | 500+ models, 80+ providers | 140+ providers, 1,800+ models (bring your own keys) | 170+ models, 16 vendors |
| Setup time | Minutes | Hours (deploy, database, keys, monitoring) | Minutes |
| Who maintains uptime | OpenRouter | You | Oxyy |
Key Takeaways
- OpenRouter never costs less than going direct. It passes through list prices and adds a 5.5% fee when you buy credits (8% on Business).
- LiteLLM is free software, not free usage. You still pay every provider's list price, plus hosting, plus engineering time.
- Oxyy is the only one of the three that lowers your per-token price: 40% of the standard rate on every model during its current launch promotion.
- LiteLLM wins on control: self-hosted, virtual keys, team budgets, guardrails and audit logs (Enterprise).
- OpenRouter wins on catalog size: 500+ models and free model variants.
- You can combine them. Put LiteLLM in front of Oxyy to keep self-hosted governance while paying 60% less per token.
What Is OpenRouter?
OpenRouter is a managed, OpenAI-compatible API that gives you 500+ models from 80+ providers through one key and one credit balance. It handles provider fallback automatically: if a provider returns an error, OpenRouter retries the next provider, and failed attempts aren't billed.
How OpenRouter makes money: inference is billed at each provider's list price on every plan, and OpenRouter charges a fee when you buy credits: 5.5% on Standard, 8% on Business, with discounted fees on Enterprise. OpenRouter says it doesn't discount inference.
What Is LiteLLM?
LiteLLM is an open-source AI gateway and Python SDK (61.7K GitHub stars) that exposes 140+ providers and 1,800+ models through one OpenAI-compatible endpoint. You deploy it on your own infrastructure and connect your own provider API keys.
Open source (free): virtual keys, users and teams, spend tracking, budgets and rate limits, load balancing, fallbacks, guardrails, logging, Prometheus metrics, an admin UI and an MCP gateway.
Enterprise (paid, annual): SSO and SCIM, OIDC/JWT auth, audit logs, secret managers, org and team admins, multi-region and support SLAs. LiteLLM prices Enterprise by annual gateway request capacity, deployment architecture and support needs, never per token.
What Is Oxyy.ai?
Oxyy.ai is a managed unified AI API: one OpenAI-compatible endpoint (https://api.oxyy.ai/v1) for 170+ models across text, vision, images, video, speech and embeddings. The OpenAI, Anthropic and Gemini SDKs all work unchanged. Requests are routed by priority and health, and when a model has more than one upstream, a failing provider is taken out of rotation and the request is retried elsewhere.
How Oxyy prices: every model shows a standard price and a charged price. A 60% "Launch Promo" is currently applied to every rate, so you pay 40% of standard, with no subscription, no seat fee and no platform fee.
Pricing: OpenRouter vs LiteLLM vs Oxyy
Cost at Different Usage Levels (List-Price Usage per Month)
| Monthly usage at list price | OpenRouter (Standard, +5.5%) | LiteLLM (self-hosted) | Oxyy.ai |
|---|---|---|---|
| $1,000 | $1,055 | $1,000 + infra | $400 |
| $10,000 | $10,550 | $10,000 + infra | $4,000 |
| $100,000 | $105,500 | $100,000 + infra + Enterprise license (if needed) | $40,000 |
Model-by-Model Price (per 1M tokens, input / output)
| Model | List price (direct / LiteLLM) | OpenRouter incl. 5.5% fee | Oxyy.ai |
|---|---|---|---|
| Claude Opus 5.5 | $4.00 / $20.00 | $4.22 / $21.10 | $1.60 / $8.00 |
| Claude Sonnet 5.5 | $2.00 / $10.00 | $2.11 / $10.55 | $0.80 / $4.00 |
| GPT-5.5 | $5.00 / $30.00 | $5.28 / $31.65 | $2.00 / $12.00 |
| Gemini 3.1 Pro | $2.00 / $12.00 | $2.11 / $12.66 | $0.80 / $4.80 |
| GLM-5.3 | $1.40 / $4.40 | $1.48 / $4.64 | $0.56 / $1.76 |
List prices from Anthropic, Z.ai and the standard prices shown on Oxyy's model pages. OpenRouter column applies its 5.5% Standard-plan credit fee to list price. Oxyy prices from oxyy.ai model pages, October 20, 2026.
The Hidden Costs of Self-Hosting LiteLLM
LiteLLM's license is free, but a production deployment isn't:
- Infrastructure: the gateway itself plus a database for keys, spend and logs, and usually a cache.
- Engineering time: upgrades, security patches, monitoring and on-call. LiteLLM ships fast (54,995 commits on GitHub), so staying current is ongoing work.
- Provider accounts: you need your own account with each provider, and you inherit each provider's rate-limit tier. For example, Anthropic's Start tier caps spend at $500 a month.
- No volume discount on tokens: you pay list price unless you negotiate contracts yourself.
Features Compared
| Feature | OpenRouter | LiteLLM | Oxyy.ai |
|---|---|---|---|
| OpenAI-compatible API | Yes | Yes | Yes |
| Anthropic Messages endpoint (for Claude Code) | Yes | Yes | Yes (https://api.oxyy.ai) |
| Google GenAI SDK compatible | Check docs | Yes (as a provider) | Yes |
| Automatic failover | Yes, across providers | Yes (you configure) | Yes, when a model has multiple upstreams |
| Budgets and spend controls | Yes (Standard and up) | Yes (per key, team, org) | Prepaid balance, per-request activity log |
| Virtual keys per team | Workspaces (5 on Standard, 1,000 on Business) | Yes | Multiple API keys |
| SSO / SCIM | Enterprise | Enterprise | — |
| Self-hosted / air-gapped | No | Yes | No |
| Free tier | 25+ free models (rate-limited) | Free software | Free starting balance + free models |
| Live per-model latency and success rate | Latency and throughput per provider | Your own metrics | Latency, throughput, TTFT and 30-day success rate on every model page |
| Price below list | No | No | Yes (60% off, current promo) |
Rate Limits Compared
- OpenRouter: no plan-based limit on paid models; limits come from the upstream provider, with fallback to other providers. Free models are capped at 20 requests per minute and 50 per day, rising to 1,000 per day after you buy $10 of credits.
- LiteLLM: you get whatever each provider gives your own account, and you can set your own limits per key or team on top.
- Oxyy: limits are set by your Oxyy tier and apply per API key (requests per minute, tokens per minute, daily caps and concurrency). Higher tiers unlock automatically as you top up, and every response carries
X-RateLimit-*headers.
Best of Both Worlds: LiteLLM in Front of Oxyy
If you need self-hosted governance (virtual keys, team budgets, guardrails, audit logs) and lower token prices, use Oxyy as an upstream inside LiteLLM. LiteLLM treats any OpenAI-compatible endpoint as a provider:
# config.yaml
model_list:
- model_name: claude-sonnet
litellm_params:
model: openai/claude-sonnet-5-5
api_base: https://api.oxyy.ai/v1
api_key: os.environ/OXYY_API_KEY
- model_name: glm
litellm_params:
model: openai/glm-5.3
api_base: https://api.oxyy.ai/v1
api_key: os.environ/OXYY_API_KEYYour teams keep calling your LiteLLM proxy as before. LiteLLM tracks spend and enforces budgets, and Oxyy bills the tokens at 40% of list price.
Switching From OpenRouter to Oxyy
Both are OpenAI-compatible, so migration is a base URL, a key and model IDs:
from openai import OpenAI
import os
client = OpenAI(
base_url="https://api.oxyy.ai/v1", # was https://openrouter.ai/api/v1
api_key=os.environ["OXYY_API_KEY"], # was OPENROUTER_API_KEY
)
resp = client.chat.completions.create(
model="claude-sonnet-5-5", # Oxyy model IDs have no vendor prefix
messages=[{"role": "user", "content": "Hello"}],
)Which One Should You Choose?
| If you… | Choose |
|---|---|
| Want the lowest per-token price without running servers | Oxyy.ai |
| Need the widest catalog, including niche open-source models and free variants | OpenRouter |
| Must self-host for compliance, or need SSO, audit logs and team chargeback | LiteLLM (Enterprise for SSO/audit) |
| Need self-hosted governance and lower token prices | LiteLLM + Oxyy as upstream |
| Already have negotiated direct provider contracts | LiteLLM with your own keys |
| Want EU or US in-region routing on a managed gateway | OpenRouter Business |
Final Verdict
OpenRouter sells convenience and catalog size. LiteLLM sells control. Neither lowers what you pay per token. Oxyy is the managed option that does: 40% of list price on every model during the current promotion, with the same OpenAI-compatible interface. For most startups and product teams, that means switching saves 60% of the AI bill with a one-line change. And if you need LiteLLM's governance, you can keep it and point it at Oxyy.
Get a free Oxyy API key and compare your own bill →
Related: Oxyy price list · Claude API pricing · OpenAI API pricing · Gemini API pricing
Frequently Asked Questions
What is the difference between OpenRouter and LiteLLM?
OpenRouter is a managed gateway you pay through credits, at provider list price plus a 5.5% credit fee. LiteLLM is free open-source software you host yourself with your own provider keys, so you pay list price plus your own infrastructure.
Is LiteLLM free?
The LiteLLM open-source gateway is free to self-host with no license fee. LiteLLM Enterprise, which adds SSO, audit logs and support SLAs, is a paid annual license priced by request capacity.
How much does OpenRouter charge?
OpenRouter bills provider list prices with no token markup and charges a fee when you buy credits: 5.5% on the Standard plan and 8% on the Business plan. BYOK usage is free up to $25,000 of list-price inference per month, then 5%.
Is OpenRouter cheaper than LiteLLM?
Not on tokens. Both bill provider list prices; OpenRouter adds a 5.5% credit fee while LiteLLM adds your hosting and maintenance costs. Which is cheaper overall depends on your volume and engineering time.
What is the cheapest unified AI API?
Oxyy.ai is the cheapest in this comparison. It currently charges 40% of the standard price on 170+ models, so $1,000 of list-price usage costs $400.
Can I use LiteLLM with Oxyy?
Yes. Add Oxyy to your LiteLLM config.yaml as an OpenAI-compatible provider with api_base https://api.oxyy.ai/v1 and your Oxyy key, then keep using LiteLLM's virtual keys and budgets.
Does OpenRouter have rate limits?
OpenRouter sets no plan-based limit on paid models; limits come from upstream providers. Free models are capped at 20 requests per minute and 50 per day, or 1,000 per day after buying $10 of credits.
Which is best for Claude Code: OpenRouter, LiteLLM or Oxyy?
All three expose an Anthropic-format endpoint. Oxyy is the cheapest, at 40% of Anthropic's list price; set ANTHROPIC_BASE_URL to https://api.oxyy.ai.
Is Oxyy self-hosted?
No. Oxyy is a managed gateway. If you need self-hosting, run LiteLLM and use Oxyy as an upstream provider.
Is the Oxyy discount permanent?
The 60% discount is labeled a "Launch Promo" on Oxyy's model pages. Each page shows the standard price and the charged price side by side.