OpenRouter alternatives: 17 gateways and model providers compared

Updated · 7 min read

The closest OpenRouter alternatives are other hosted gateways that sell Claude, GPT, Gemini and open models under one key: Vercel AI Gateway, Requesty, Kilo Gateway, Helicone, Cloudflare AI Gateway, NanoGPT, Eden AI and AIML API. If you only use open models, providers such as Together AI, Fireworks, DeepInfra, Groq and Cerebras host them directly. If you want no middleman, run LiteLLM or Bifrost yourself in front of your own provider keys. For the lowest fees on Claude, GPT and Gemini, look at Vercel AI Gateway or Kilo Gateway, which add no markup; for the closest like-for-like swap, Requesty. All of them have an OpenAI-compatible API, so in most BYOK apps you switch by changing the base URL, the key and the model name.

What OpenRouter charges, for comparison

OpenRouter passes model prices through without markup and charges a fee when you buy credits: 5.5% with a $0.80 minimum (5% with crypto, 8% on its Business plan). Its :free models allow 50 requests a day, or 1,000 once you have bought $10 of credits, and using your own provider keys is free up to $25,000 a month, then 5%. Setup is in the OpenRouter API key guide, and the current free list is in OpenRouter free models.

Gateways compared

Checked on each vendor's pricing and docs pages on 2026-11-18. "Closed" means models such as Claude, GPT and Gemini.

Gateway Models Added to model prices Free use Your own provider keys OpenAI-compatible base URL
OpenRouter (for comparison) 600+, closed and open 5.5% on credit purchases, $0.80 minimum :free models, 50 requests/day (1,000 after $10 of credits) Free up to $25,000/month, then 5% https://openrouter.ai/api/v1
Vercel AI Gateway 400+, closed and open Nothing (payment processing fees may apply) Monthly credit for some models Yes, no markup; needs purchased credits https://ai-gateway.vercel.sh/v1
Requesty 600+, closed and open 5% on each top-up Free models, 200 requests/day, no card OpenAI, Anthropic, Google, Mistral, xAI https://router.requesty.ai/v1
Kilo Gateway Closed and open Nothing, and no payment fee for now :free models, 200 requests/hour per IP Yes, $0 on Kilo's side https://ai-gateway.kilo.ai/api/v1
Helicone AI Gateway GPT, Claude, Gemini and 100+ more 0% markup on credits Hobby plan, 10,000 requests/month Yes https://ai-gateway.helicone.ai
Cloudflare AI Gateway Workers AI, OpenAI, Anthropic, Google, xAI and more 5% on Unified Billing credits Gateway features are free Yes, stored in the gateway https://api.cloudflare.com/client/v4/accounts/<ACCOUNT_ID>/ai/v1
NanoGPT Closed and open No fee on deposits; +5% to pin a provider One free model, web only Yes, 5% of the normal price https://api.nano-gpt.com/api/v1
Eden AI Closed and open 5.5% platform fee at checkout None listed On plans that include it https://api.edenai.run/v3
AIML API Closed and open No published fee Free tier paused Not documented https://api.aimlapi.com/v1
Portkey The providers you connect Free Developer plan or $49/month 10,000 logs/month Required https://api.portkey.ai/v1
LiteLLM (self-hosted) 100+ providers Nothing (open source) Free to run Required http://localhost:4000
Bifrost (self-hosted) 20+ providers Nothing (open source) Free to run Required http://localhost:8080/v1

Hosted gateways, one by one

Vercel AI Gateway

Vercel charges the provider's list price with no markup or platform fee, including on your own keys. Every Vercel team gets a monthly included credit that covers a subset of models at lower rate limits; once you buy credits, that monthly credit stops. Your own keys need purchased credits, and by default a failed request falls back to Vercel's keys and is charged to your balance.

Requesty

Requesty works most like OpenRouter. It charges provider prices plus a 5% fee on top-ups, and model IDs are either provider/model or a bare name that Requesty routes with failover. Free models need no card. It also has an Anthropic-style endpoint for tools such as Claude Code.

Kilo Gateway

Kilo runs this gateway for its Kilo Code coding agent, and any OpenAI-compatible app can use it. Usage is billed at the upstream provider's rate, and Kilo says it is not currently charging a payment-processing fee on credits.

Helicone

Helicone is an observability platform with a gateway. You can buy credits at 0% markup, or attach your own keys and use it for logs and cost tracking. Paid plans start at $79 a month.

Cloudflare AI Gateway

Cloudflare's gateway is free for analytics, caching and rate limiting. You either store your own provider keys in the gateway, or prepay through Unified Billing (5% fee on credit purchases, no markup on tokens). The OpenAI-compatible endpoint needs your account ID in the URL and a Cloudflare API token, so it suits developers more than chat apps.

NanoGPT

NanoGPT bills text at list price with no percentage fee on deposits, which start at $1 by card. Its PRO plan ($12 a month) includes 60M input tokens a week on open models such as GLM, DeepSeek, Kimi and Qwen; Claude, GPT, Gemini and Grok are billed from your balance. NanoGPT's own migration guide warns that some models cost more there than on OpenRouter even after OpenRouter's fee, so compare the models you use.

Eden AI

Eden AI charges the same 5.5% as OpenRouter, at checkout and with no markup on models, and uses provider/model IDs such as openai/gpt-latest. Its base URL ends in /v3, not /v1.

AIML API

AIML API lists its own per-model prices. Some match the provider's price and some are higher. On 2026-11-18 it charged $3.174 / $15.87 per 1M tokens for Claude 4.5 Sonnet, against $3 / $15 at Anthropic. Its FAQ says the free tier is paused.

Portkey

Portkey does not sell model access. You add your provider credentials and call them as @provider/model, and Portkey adds budgets, rate limits and logs for teams on top.

Self-hosted: LiteLLM and Bifrost

The LiteLLM proxy is an MIT-licensed server that puts 100+ providers behind one OpenAI-compatible endpoint. You give each app or person a virtual key with its own models, budget and rate limit, and you pay each provider directly. You run and secure the server yourself, and apps on your phone can only use it if they can reach that server. SSO and audit logs are in the paid Enterprise edition.

Bifrost is an Apache-2.0 gateway that starts with npx -y @maximhq/bifrost or Docker and serves 20+ providers on port 8080. Both projects had new releases in November 2026.

Providers that host many open models

These companies run the models on their own hardware, so one key covers their whole catalog, mostly open-weight models. Hugging Face instead routes each request to a partner provider; its docs say the routing is powered by OpenRouter.

Provider Models How you pay Free use OpenAI-compatible base URL
Hugging Face Inference Providers Open models from partner providers Provider prices, no markup PRO accounts get $2/month https://router.huggingface.co/v1
Together AI Open models Prepaid, $15 minimum first purchase No free trial https://api.together.ai/v1
Fireworks AI Open models Prepaid credits None listed https://api.fireworks.ai/inference/v1
DeepInfra Open models, plus Claude and Gemini Card or prepaid balance None listed https://api.deepinfra.com/v1/openai
Groq gpt-oss, Qwen Monthly invoice on the Developer tier Free plan, 1,000 requests/day for gpt-oss https://api.groq.com/openai/v1
Cerebras gpt-oss, Qwen, Kimi, GLM, Gemma Prepaid credits $5 trial credit for 30 days https://api.cerebras.ai/v1

On 2026-11-18 gpt-oss-120b cost $0.15 input and $0.60 output per 1M tokens on Together, Fireworks and Groq alike. DeepInfra also lists Claude and Gemini at the providers' own prices, for example Claude Sonnet 5.5 at $2 / $10. Hugging Face gives free accounts no monthly credit, and Cerebras says it has no permanent free tier. Apps that accept these keys are listed on the Together AI, Fireworks AI, Cerebras and Hugging Face pages.

Switching an app from OpenRouter

  1. Create a key at the new service and set a spend limit if it offers one.
  2. In the app, add a custom OpenAI-compatible provider and paste the base URL from the tables above and your new key.
  3. Pick your models again. IDs differ between services, and OpenRouter's :free suffix does not carry over.
  4. If an app offers only OpenRouter and no custom endpoint, keep an OpenRouter key for that app.

Coding agents that speak the Anthropic Messages API, such as Claude Code, need an Anthropic-compatible endpoint instead; Requesty and Cloudflare document one.

Gateways and apps that take an OpenRouter key

  • OpenRouter

    Pay as you go; BYOK usage has a small fee

    Visit

    Unified API and marketplace for top AI models (OpenAI, Anthropic, Google, Mistral, Cohere, Meta), with routing, transparent pricing, and flexible billing/BYOK options.

    Works with:OpenAIAnthropicGoogle GeminiMistral+2 more

  • Kilo Code

    Free to start, no markup on model usage · VS Code, JetBrains, CLI · Open source

    Visit

    Open-source AI coding agent for VS Code, JetBrains and the CLI. Use 500+ models at zero markup or bring your own keys.

    Works with:AnthropicOpenAIGoogle GeminiOpenRouter+8 more

  • Infinity

    Free · macOS

    Visit

    Free Mac AI chat app with git-style branching: fork any part of an answer into its own thread, ask up to four models at once, and use your own keys or local models.

    Works with:OpenAIAnthropicGoogle GeminixAI+7 more

  • Caotang

    Free · Web · Open source

    Visit

    Free, local-first browser writing studio for novels and scripts, with an AI agent that runs on your own Anthropic, OpenAI, Google, OpenRouter or local model key.

    Works with:AnthropicOpenAIGoogle GeminixAI+5 more

  • Shizue

    Free · Chrome extension · Open source

    Visit

    Free, open-source Chrome side panel for AI chat, page summaries, bilingual video captions and in-tab tasks, running on your ChatGPT plan or your own OpenAI, Claude, Gemini or OpenRouter key.

    Works with:OpenAIAnthropicGoogle GeminiOpenRouter+1 more

  • PockoBots

    Free (beta) · Android · Open source

    Visit

    Open-source AI coding agent and personal bots that run on your Android phone, powered by your own OpenAI, Anthropic, Gemini, Groq or OpenRouter key with no app subscription.

    Works with:OpenAIAnthropicGoogle GeminiGroq+1 more

  • Conduit

    Free · Windows, macOS, Linux · Open source

    Visit

    Open-source desktop AI app for agents, research, writing and slides that runs on your own keys from 23 providers or local models.

    Works with:AnthropicOpenAIGoogle GeminixAI+12 more

  • SoloMD

    Free · macOS, Windows, Linux · Open source

    Visit

    Free, open-source Markdown editor for desktop and mobile with built-in AI tools that run on your own key from 14 providers.

    Works with:OpenAIAnthropicGoogle GeminixAI+7 more

See all 46 apps that work with OpenRouter

Which should you pick

  • Stay on OpenRouter if you rely on its free models or on apps that only support OpenRouter. The fee is 5.5%, or $0.80 on a $10 top-up because of the minimum.
  • Move to Vercel AI Gateway, Kilo Gateway or Helicone if you mostly pay for Claude, GPT or Gemini and want no markup.
  • Pick Requesty if you want OpenRouter-style routing and free models with a slightly lower fee.
  • Use Groq, Together AI or Hugging Face if you only run open models.
  • Self-host LiteLLM if you already have provider keys and want one endpoint with per-app budgets.
  • Go direct if you mainly use one provider. You skip every gateway fee, you get features a gateway may not pass through yet, and only a direct key gets the provider's own free tier, such as Gemini's: see the Claude, OpenAI, Gemini and DeepSeek key guides.

Questions

What is the best OpenRouter alternative?

It depends on what you use OpenRouter for. For Claude, GPT and Gemini with no markup, try Vercel AI Gateway or Kilo Gateway; for an OpenRouter-style router with free models, Requesty; for your own provider keys behind one endpoint, self-hosted LiteLLM.

Is there an OpenRouter alternative without fees?

Yes. On 2026-11-18 Vercel AI Gateway, Kilo Gateway and Helicone charged no markup on model usage, and NanoGPT charged no percentage fee on deposits. Requesty adds 5% to top-ups, Cloudflare 5% to Unified Billing credits and Eden AI 5.5%, close to OpenRouter's 5.5%.

OpenRouter vs LiteLLM: what is the difference?

OpenRouter is a hosted service: you buy credits from it and it pays the model providers. LiteLLM is open-source software you run yourself in front of your own provider keys, so there is no fee, but you host and secure the server.

OpenRouter vs Requesty: which is cheaper?

Both pass model prices through. Requesty adds 5% to each top-up, while OpenRouter charges 5.5% with a $0.80 minimum, so small top-ups cost relatively more on OpenRouter. Requesty's free models allow 200 requests a day; OpenRouter's allow 50, or 1,000 after you buy $10 of credits.

Is there a free alternative to OpenRouter?

Groq has a free plan for open models such as gpt-oss-120b (1,000 requests a day), Requesty and Kilo Gateway have free models, and Vercel AI Gateway includes a monthly credit for some models. For a free Gemini, use Google's own free tier.

OpenRouter vs Together AI?

Together AI runs open models such as DeepSeek, Qwen, Kimi, GLM and gpt-oss on its own servers; Claude, GPT and Gemini chat models are not on its pricing page. It is fully prepaid, with a $15 minimum first purchase and no free trial.

Sources

Checked against the providers' own documentation on 2026-11-18. Prices and limits change; the links below have the current figures.

Related guides