OpenRouter Alternatives in 2026
Short version: there are three ways to reach Claude, GPT, Gemini and other models through one integration. A hosted gateway (OpenRouter, Requesty, Vercel AI Gateway, byesu) gives you one key and one bill. A self-hosted or bring-your-own-key gateway (LiteLLM, Portkey, Cloudflare AI Gateway) puts routing, logging and budgets in front of provider accounts that you own. Going direct to each provider removes the middle layer and leaves the integration work to you. None of these is the best choice for everyone; the right one depends on who should hold the provider keys, who should run the infrastructure and which API formats your clients speak.
Last verified: 2026-11-14. Every statement about a product other than byesu comes from that product's own site or documentation, read on that date and listed under Sources. Where an official page did not state something, this page leaves it out. byesu publishes this page, so read the byesu entry with that in mind.
Comparison Table
| Option | Type | API formats | Billing model | Bring your own provider keys | Self-hostable | Public status / latency data |
|---|---|---|---|---|---|---|
| OpenRouter | Hosted aggregator | OpenAI Chat Completions, OpenAI Responses, Anthropic Messages-compatible input | Prepaid credits; inference at provider list price, platform fee on credit purchases | Optional | Hosted service | Not confirmed on the pages we read |
| Requesty | Hosted gateway | OpenAI Chat Completions, OpenAI Responses, Anthropic Messages | Prepaid USD balance; service fee added to each top-up | Optional | Hosted service | Not confirmed on the pages we read |
| Vercel AI Gateway | Hosted gateway | OpenAI Chat Completions, OpenAI Responses, Anthropic Messages, AI SDK | Prepaid credits at provider list price, no markup on tokens | Optional | Hosted service | Not confirmed on the pages we read |
| byesu | Hosted gateway | OpenAI Chat Completions, OpenAI Responses, Anthropic Messages | Prepaid balance, pay-as-you-go by token usage; live price in the console | Not offered | Not offered | Yes: status page and weekly benchmarks |
| LiteLLM | Open-source SDK and self-hosted proxy | OpenAI Chat Completions, OpenAI Responses, Anthropic Messages and others | You pay providers directly; enterprise features need a license key | Yes, that is the model | Yes | Not applicable (you run it) |
| Portkey | Open-source gateway with a managed cloud | Universal API, usable from the OpenAI SDK | Plans based on recorded logs; provider credentials are yours | Yes | Yes (open-source gateway) | Not confirmed on the pages we read |
| Cloudflare AI Gateway | Managed proxy on Cloudflare | Provider-native formats, plus an OpenAI-compatible chat completions endpoint | Core features free; optional Unified Billing credits with a fee on purchases | Yes, or use Unified Billing | No (runs on Cloudflare) | Not confirmed on the pages we read |
| Direct provider APIs | No gateway | Each provider's own format | One account and one bill per provider | Not applicable | Not applicable | Each provider's own status page |
"Not confirmed on the pages we read" means only that: the official pages checked for this article did not state it. It is not a claim that the data does not exist.
Hosted Aggregators and Gateways
You get one API key and one balance; the service holds the provider relationships.
OpenRouter (the baseline)
What it is. OpenRouter describes itself as access to hundreds of AI models through a single API endpoint, with automatic fallbacks. Its pricing page lists 600+ models from 80+ providers on the paid plans (checked 2026-11-14). The API reference documents POST /api/v1/chat/completions and POST /api/v1/responses, and its Claude Code guide says OpenRouter exposes an input compatible with the Anthropic Messages API.
Who it suits. Teams that want the widest catalog, provider routing they can tune (sort by price, throughput or latency, with fallbacks on by default), and the option to attach their own provider keys.
What to check. Per the pricing page (checked 2026-11-14), inference is billed at each provider's list price and a platform fee applies when you buy credits: 5.5% on Standard and 8% on Business. With your own keys, the first $25,000 of list-price inference per month carries no fee and 5% applies after that. For Claude Code, the guide recommends putting the Anthropic first-party provider first for maximum compatibility. A longer side-by-side is in byesu vs OpenRouter.
Requesty
What it is. A hosted gateway whose docs describe it as one OpenAI-compatible API for 600+ models from 30+ providers. The base URL is https://router.requesty.ai/v1, and the API reference lists three formats: OpenAI Chat Completions, the OpenAI Responses API and Anthropic Messages. An EU endpoint keeps processing and storage in Frankfurt.
Who it suits. Teams that want a hosted router with policy features in the same product: fallback, load balancing and latency routing policies, usage analytics, budgets per team, user and key, guardrails and role-based access.
What to check. Requesty bills from a prepaid USD balance, and its billing page says a 5% service fee is added to every top-up (checked 2026-11-14). Bring-your-own-key is documented for a fixed list of providers, with at most one key per provider.
Vercel AI Gateway
What it is. A hosted API at https://ai-gateway.vercel.sh. The docs say your application does not need to run on Vercel; you need a Vercel account to create a key and manage billing. It can be called through the AI SDK, OpenAI Chat Completions, the OpenAI Responses API or Anthropic Messages.
Who it suits. Teams already on Vercel or the AI SDK, and teams that want request logs, budgets, provider failover and bring-your-own-key without operating their own proxy.
What to check. The pricing page says AI Gateway charges no markup and no platform fee on tokens, that you pay the provider's list price from prepaid credits, and that you are responsible for any payment processing fees (checked 2026-11-14). Some capabilities are paid add-ons, bring-your-own-key needs purchased credits, and budgets do not count bring-your-own-key spend. The free tier covers a subset of models.
byesu
What it is. An AI API gateway with two endpoints: OpenAI-compatible at https://byesu.com/v1 (/v1/chat/completions and /v1/responses) and Anthropic-native at https://byesu.com (/v1/messages). One key covers Claude, GPT, Gemini, Grok and other models.
Who it suits. Individuals and small teams who want one key without running anything, especially for coding tools: there are setup guides for Claude Code, Codex CLI and Cherry Studio.
What to check. Billing is pay-as-you-go by token usage with no subscription; the live price for each model is shown in the console after sign-in, and the console records usage per request. The catalog is curated and smaller than OpenRouter's. byesu does not offer bring-your-own-key or a self-hosted version, so it is not the right fit if you need to use your own provider contracts. Availability and latency are public on the status page and the weekly benchmarks page. See Pricing & Billing for how charges are calculated.
Self-Hosted and Bring-Your-Own-Key Gateways
Here you keep your own provider accounts. The gateway adds a common interface, routing, logging and spend controls in front of them.
LiteLLM
What it is. An open-source project with two parts: a Python SDK that calls 100+ LLMs through one interface, and a proxy server (AI Gateway) that its docs describe as a self-hosted gateway for platform teams. The proxy exposes /chat/completions, /responses, /messages and other endpoints.
Who it suits. Teams that want prompts and keys to stay in their own infrastructure and want virtual keys, per-team budgets and rate limits, spend tracking, fallbacks and load balancing under their own control.
What to check. You deploy and operate it (Docker, Helm or Terraform). Virtual keys and budget enforcement require a Postgres database. SSO, audit logs and some other features sit behind an enterprise license key. LiteLLM can also sit in front of a hosted gateway; see byesu vs LiteLLM.
Portkey
What it is. An AI gateway that its docs describe as a unified interface to over 250 AI models, with observability, caching, fallbacks, load balancing, conditional routing, retries, guardrails, budget limits and rate limits. The gateway is open source under the MIT license and starts locally with npx @portkey-ai/gateway; Portkey also runs a managed cloud version and offers private deployments for enterprises.
Who it suits. Teams that want gateway controls plus observability, with the choice between a managed service and running the open-source gateway themselves.
What to check. You add your own provider credentials in the Model Catalog and Portkey stores them. The pricing page describes plans by recorded logs per month, with a free developer tier (checked 2026-11-14). The official pages we read did not describe inference being billed through Portkey, so plan on paying your providers directly.
Cloudflare AI Gateway
What it is. A managed proxy on Cloudflare's network for analytics, logging, caching, rate limiting and routing with fallbacks. You connect to providers in their native API formats, and there is an OpenAI-compatible chat completions endpoint that takes {provider}/{model} names.
Who it suits. Teams that already use Cloudflare and mainly want visibility and control over calls to provider accounts they already have.
What to check. The pricing page says the core features are offered for free (checked 2026-11-14). You can store provider keys in the gateway, or use Unified Billing to call supported providers without your own keys by buying credits; per the same page a 5% fee applies to credits purchased through Unified Billing, and provider inference pricing is passed through with no markup. It is not self-hostable. Log storage has its own limits and pricing.
Going Direct to Providers
What it is. Separate accounts and API keys with each model provider, with no gateway in between.
Who it suits. Teams that use one or two providers, need provider-specific features on the day they ship, or want a direct contract and support relationship with the provider.
What to check. Each provider has its own API format, billing account, rate limits and status page. Using several of them means integrating and monitoring each one, and writing your own fallback logic if you need it. A library such as the LiteLLM SDK can reduce the code side of that work without adding a network hop.
How to Choose
- Decide who holds the provider keys. If you need your own contracts, credits or data agreements, look at LiteLLM, Portkey, Cloudflare AI Gateway or going direct, or at the bring-your-own-key options of OpenRouter, Requesty and Vercel AI Gateway. If you want one key and one balance, look at the hosted gateways.
- Decide who runs the infrastructure. Self-hosting gives control and takes operations work (a database, upgrades, monitoring). A hosted gateway removes that work.
- Match the API format to your clients. Claude Code needs an Anthropic Messages endpoint; Codex CLI needs the OpenAI Responses API; most other tools use OpenAI Chat Completions. Check the table above.
- Read the fee structure, not just the token price. Some services add a fee when you buy credits, some charge for add-ons or logs, and some add nothing on top of provider prices. Check the current pricing page of each.
- Measure from where your code runs. Published numbers are a starting point. How to measure LLM API latency shows a way to test any endpoint yourself.
- Keep the exit cheap. All of the gateways here accept OpenAI-style requests, so switching is mostly a base URL and a key. Avoid building on features you cannot replace.
Related Links
- byesu vs OpenRouter — detailed two-way comparison
- byesu vs LiteLLM — hosted gateway versus self-hosted proxy
- Quick Start (3 steps) — key, base URL, model
FAQ
What are the main alternatives to OpenRouter?
They fall into three groups. Hosted gateways such as Requesty, Vercel AI Gateway and byesu give you one key and one balance. Self-hosted or bring-your-own-key gateways such as LiteLLM, Portkey and Cloudflare AI Gateway add routing, logging and budgets in front of provider accounts you own. The third option is calling each provider directly. Which one fits depends on who should hold the provider keys and who should run the infrastructure.
Which OpenRouter alternatives can be self-hosted?
LiteLLM is an open-source SDK and proxy that you deploy yourself with Docker, Helm or Terraform. Portkey publishes its gateway as open source under the MIT license and also offers a managed cloud version. OpenRouter, Requesty, Vercel AI Gateway and byesu are hosted services, and Cloudflare AI Gateway runs on Cloudflare (checked 2026-11-14).
Which gateways work with Claude Code?
Claude Code needs an endpoint in the Anthropic Messages format. As of 2026-11-14, OpenRouter documents an Anthropic Messages-compatible input, and Requesty, Vercel AI Gateway, LiteLLM and byesu document an Anthropic Messages endpoint. On byesu you set ANTHROPIC_BASE_URL to https://byesu.com. Check each product guide for model and provider limits before relying on it.
Do these gateways add fees on top of provider prices?
It varies, so read each pricing page. As checked on 2026-11-14: OpenRouter bills inference at provider list price and charges a platform fee when you buy credits (5.5% on Standard, 8% on Business); Requesty adds a 5% service fee to every top-up; Vercel AI Gateway states no markup and no platform fee on tokens; Cloudflare applies a 5% fee to credits bought through Unified Billing. byesu bills pay-as-you-go by token usage and shows the live price in the console after sign-in.
Is byesu a drop-in replacement for OpenRouter?
For OpenAI-style clients the switch is a base URL (https://byesu.com/v1) and a key, and model names must match the byesu console list. It is not equivalent in scope: OpenRouter has a much larger catalog and supports bring-your-own-key, which byesu does not offer. byesu suits people who want one key for Claude, GPT, Gemini and Grok with pay-as-you-go billing and public status and benchmark pages.
Sources
All pages read on 2026-11-14.
- OpenRouter: quickstart, API reference overview, provider routing, Claude Code guide, pricing
- Requesty: docs home, inference APIs, billing, bring your own keys
- Vercel AI Gateway: overview, pricing
- LiteLLM: docs home, proxy overview, Docker quick start, enterprise, GitHub repository
- Portkey: what is Portkey, AI Gateway, virtual keys / Model Catalog, pricing, gateway repository
- Cloudflare AI Gateway: overview, pricing, unified API, bring your own keys, Unified Billing, providers
- byesu: status, benchmarks, pricing & billing