OpenRouter Alternatives for Indie Developers (2026 Fees)

Short answer: If you want OpenRouter's convenience without its 5.5% credit fee, Vercel AI Gateway is the closest swap: one API key, models across many providers, fallbacks, and no markup or platform fee on tokens. If you already hold provider keys, Cloudflare AI Gateway is free in front of them, and LiteLLM is the pick if you want to self-host. If you only use one provider, skip the gateway and call it directly.

Every price, fee and limit below comes from each vendor's own pricing page or docs, checked on 14 Nov 2026. The advice in between is ours and is labelled.

TL;DR#

• OpenRouter's real cost is the credit fee, not the tokens. Inference is at provider list price with no markup, but buying credits costs 5.5% ($0.80 minimum) on Standard and 8% on Business (OpenRouter pricing, OpenRouter FAQ).
• BYOK on OpenRouter is fee-free up to $25,000 of list-price inference a month, then 5% (OpenRouter pricing).
• Vercel AI Gateway: "no markup and no platform fee on tokens", including BYOK, usable from any cloud (Vercel docs).
• Cloudflare AI Gateway: core features free; Unified Billing adds 5% on credits (Cloudflare docs).
• LiteLLM: MIT-licensed proxy you host yourself; enterprise is custom-priced (LiteLLM licence, LiteLLM Enterprise).
• Portkey: free Developer plan, Production $49/month (Portkey pricing). Requesty: free plan, then a 5% markup (Requesty pricing).

Disclosure: ZeroRaised publishes this guide and is a launch site for bootstrapped products. ZeroRaised has no commercial relationship with any vendor named here and earns nothing if you sign up. Facts come from the linked sources as checked on 14 Nov 2026.

What this post covers (and what it does not)#

This is a buying guide for the gateway layer only. For the concepts, see what LLM routing is; to pick a provider, see how to choose an LLM API; and if your token bill is the real problem, how to reduce LLM costs will save more than any switch here. Helicone and Langfuse are in our LLM observability tools roundup.

What OpenRouter charges today#

OpenRouter is a hosted gateway with one API and one credit balance in front of many providers. Its pricing page lists four plans (OpenRouter pricing):

The details that matter for an indie budget:

• The fee is on credit purchases, not requests. The FAQ says: "OpenRouter charges a 5.5% ($0.80 minimum) fee when you purchase credits." Crypto payments are charged 5% (OpenRouter FAQ).
• No markup on tokens. The FAQ says: "We never mark-up the pricing of the underlying providers."
• BYOK. On Standard and Business, $25,000 of list-price inference per month carries no OpenRouter fee. Above that, a 5% fee applies, calculated on what the request would have cost at OpenRouter's list price and deducted from your credits (OpenRouter pricing).
• Free models. 20 requests per minute and 50 requests per day, rising to 1,000 per day once you have bought at least $10 of credits. Failed attempts count toward the cap (OpenRouter pricing).
• Fallbacks. You pass a models array in priority order and pay the rate of whichever model actually served the request. A :free model that falls back to a paid one is billed at the paid price (OpenRouter docs).
• Caching. Prompt caching works on supported providers, and OpenRouter routes follow-up requests to the same provider to keep the cache warm (OpenRouter docs).

Our read: if the 5.5% is your reason for leaving, try BYOK first. With your own provider keys, the fee disappears at any indie scale.

The alternatives at a glance#

Vercel AI Gateway#

Vercel's docs open the pricing page with the line that matters: "AI Gateway charges no markup and no platform fee on tokens." You pay the provider's list price from a balance of AI Gateway Credits (Vercel pricing docs).

• Not tied to Vercel hosting. You need a Vercel account for the key and billing, "but not a Vercel project or deployment". It speaks OpenAI Chat Completions, OpenAI Responses and Anthropic Messages (Vercel docs).
• Free tier. A monthly included credit, "not an expiring trial", on a subset of models. Buying credits ends the monthly free credit (Vercel FAQ). The docs did not state its dollar value, so we do not quote one.
• BYOK. No markup, but you need purchased credits and a positive balance. If a request with your key fails, Vercel retries with its own keys and charges your credits (Vercel pricing docs).
• Watch-outs. "You're responsible for any payment processing fees that may apply." Add-ons cost extra (Regional Hosting is 5% of inference cost). Purchased credits expire one year after purchase.
• Caching and failover. Prompt caching with caching: 'auto' plus provider and model fallbacks, but no full-response cache.

Our read: the most direct OpenRouter replacement for a solo developer who wants one key and one balance without a percentage fee. Leave add-ons off, and note that BYOK spend is not covered by Vercel's budgets.

Cloudflare AI Gateway#

Cloudflare's gateway sits in front of providers you already use. Its pricing page says: "AI Gateway's core features available today are offered for free", and those core features include analytics, caching and rate limiting (Cloudflare pricing). Its overview lists request retries and model fallback, and works with Workers AI, Anthropic, Google Gemini, OpenAI, Replicate and others (Cloudflare docs).

• BYOK is the free path. Pass your provider key on the request, or store one under the default alias in the gateway (Cloudflare Unified Billing docs).
• Unified Billing is the paid path. One Cloudflare bill for OpenAI, Anthropic, Google AI Studio, Google Vertex AI, xAI and Groq. "A 5% fee is applied to all credits purchased through Unified Billing", with provider pricing passed through at no markup.
• Logs. Accounts whose first gateway was created on or after 24 September 2026 store logs in Workers Logs, under that product's pricing.

Our read: if you already pay each provider directly, this is the cheapest hosted layer on the list. Unified Billing costs about what OpenRouter does, so use it only if one bill is worth 5% to you.

LiteLLM (self-hosted)#

LiteLLM is an open source proxy you run yourself. Its licence file puts everything outside the litellm/enterprise/ directory under the MIT licence (LiteLLM licence). You bring every provider key, so there is no gateway fee, only the server and your time.

• Reliability. "Fallbacks are how LiteLLM does automatic failover", with num_retries per model (LiteLLM docs).
• Caching. Exact-match caching on Redis, S3, GCS, in-memory or disk, and semantic caching on Redis, Valkey or Qdrant. The docs recommend Redis in production, because in-memory and disk caches are not shared between workers (LiteLLM caching docs).
• Enterprise. Custom pricing with a 30-day trial key, adding SSO, audit logs, RBAC and SLA support (LiteLLM Enterprise).

Our read: a good fit if you already run a VPS and want full control over keys, logs and data. The cost is operational: if it goes down, so does every AI feature in your app.

Portkey#

Portkey offers a hosted gateway with observability, plus an open source gateway you can self-host. Its plans (Portkey pricing):

Fallbacks, load balancing and retries are listed on every plan. Simple caching is listed on Production and semantic caching on Enterprise. You connect your own provider keys through its key management. The pricing page footer says Portkey is now part of Prisma AIRS AI Gateway, so check the plan details again before committing.

Our read: Portkey suits a developer who wants gateway plus prompt management and logs in one product. For a tiny app, the free plan's 3-day log retention is the limit you will feel first.

Requesty#

Requesty is the closest like-for-like hosted router to OpenRouter. Its pricing page lists a Free plan ($0, free models only, 200 requests per day, routing, caching and fallbacks, EU data residency, no card) and Pay as you go at a "5% markup" on model cost, with 600+ models and BYOK. Its own example: a model that costs $10 per 1M tokens from OpenAI costs $10.50 through Requesty (Requesty pricing).

Our read: on fees alone, Requesty's 5% is close to OpenRouter's 5.5%, so moving between them is mostly a choice about features, EU data residency and support, not price.

Direct open-model hosts: Together AI, DeepInfra, Groq#

If the main reason you use OpenRouter is cheap access to open models, you can go straight to a host and skip the gateway. Sample serverless prices per 1M tokens, input / output, as listed today:

DeepInfra's pricing page says you must "add a card or pre-pay" to use the service, with "no long-term contracts or upfront costs". Groq's rate limits docs list a Free plan and a Developer plan, with the Developer base limits at 30 requests per minute and 1K requests per day for models such as openai/gpt-oss-120b (Groq rate limits). Groq's pricing page did not show per-token prices to our fetch today, so we do not quote them.

Our read: one host plus a fallback to a second host, written as a short try/catch in your code, covers most open-model apps. You lose OpenRouter's single bill but keep every cent of the 5.5%.

Going direct to provider APIs#

The cheapest gateway is no gateway. On one provider, its own SDK gives you list price, new features first and one fewer hop. You lose cross-provider fallback and one bill, but at small scale a retry and an error alert is often enough.

Worked cost example#

This is an illustrative example using the fees above, not a quote. Say your app spends $500 a month on inference at provider list prices, all on paid models.

At $500 a month, the spread between hosted options is under $50. Our read: pick on features and reliability, because a cheaper model saves far more than a cheaper gateway.

Picks by situation#

How to switch in an afternoon#

1. List what you use: models, fallbacks, streaming, tool calls and OpenRouter-specific options.
2. Change the base URL and key. Where the target accepts OpenAI-style requests (Vercel's docs list it), this is mostly config plus model-name mapping; check each gateway's docs for the rest.
3. Recreate fallbacks in the new gateway's syntax.
4. Shadow some traffic and compare errors, latency and cost before switching fully.
5. Set a budget or alert, on the gateway or on the provider if you use your own keys.

Once the AI features are stable, run through our product launch checklist for bootstrapped makers before you ship.

Frequently asked questions#

Does OpenRouter mark up model prices?#

No. OpenRouter's FAQ says it never marks up the pricing of the underlying providers. It makes its money on credit purchases instead: 5.5% with a $0.80 minimum on the Standard pay-as-you-go plan, 8% on Business, and 5% for crypto payments. With BYOK, the first $25,000 of list-price inference each month carries no fee, then 5% applies.

What is the cheapest OpenRouter alternative?#

Going direct to one provider's API has no gateway overhead at all. Among gateways, Vercel AI Gateway charges no markup or platform fee on tokens, Cloudflare AI Gateway is free when you use your own provider keys, and a self-hosted LiteLLM proxy costs only the server you run it on. Which is cheapest in practice depends on whether you value your own time more than a few percent of your token bill.

Is there a free OpenRouter alternative?#

Several. Cloudflare AI Gateway's core features are free, Portkey's Developer plan is free with 10K recorded logs a month, Requesty's Free plan allows 200 requests a day on free models, Vercel AI Gateway includes a monthly free credit on a subset of models, and LiteLLM is open source. You still pay the model providers for tokens unless you stick to free models.

Should I use a gateway at all or call providers directly?#

Our read: if your app uses one provider and one or two models, call it directly and skip the extra hop. A gateway earns its keep when you need fallbacks across providers, want to try many models without managing many accounts, or need one place for logs and budgets.

Can I switch away from OpenRouter without rewriting my app?#

Usually, yes. Vercel AI Gateway's docs, for example, say you can keep an existing OpenAI or Anthropic SDK and change its base URL, so a move is often a base URL, an API key and model name changes. Check model naming, fallback syntax and any provider-specific options, then run a small share of traffic through the new gateway before switching fully.

Launch your AI app#

ZeroRaised is a launch directory for products built without institutional VC. When your AI app is ready, submit it to ZeroRaised for free. For more guides on building and growing without investors, join the ZeroRaised newsletter.

Sources#

• OpenRouter pricing, last checked 14 Nov 2026
• OpenRouter FAQ, last checked 14 Nov 2026
• OpenRouter model fallbacks docs, last checked 14 Nov 2026
• OpenRouter prompt caching docs, last checked 14 Nov 2026
• Vercel AI Gateway pricing docs, last checked 14 Nov 2026
• Vercel AI Gateway overview, last checked 14 Nov 2026
• Vercel AI Gateway FAQ, last checked 14 Nov 2026
• Cloudflare AI Gateway pricing, last checked 14 Nov 2026
• Cloudflare AI Gateway overview, last checked 14 Nov 2026
• Cloudflare AI Gateway Unified Billing, last checked 14 Nov 2026
• LiteLLM licence (GitHub), last checked 14 Nov 2026
• LiteLLM Enterprise, last checked 14 Nov 2026
• LiteLLM proxy reliability docs, last checked 14 Nov 2026
• LiteLLM proxy caching docs, last checked 14 Nov 2026
• Portkey pricing, last checked 14 Nov 2026
• Requesty pricing, last checked 14 Nov 2026
• DeepInfra pricing, last checked 14 Nov 2026
• Together AI pricing, last checked 14 Nov 2026
• Groq rate limits docs, last checked 14 Nov 2026