OpenRouter Alternatives: Fees and Same-Model Prices

Two separate things decide what a gateway costs you: what it charges for a model, and what it charges on top. People comparing OpenRouter alternatives usually look at only one of them. This page covers both, and where Openference is not the cheaper choice it says so.

What OpenRouter charges on top

OpenRouter describes itself as "One API for 600+ models from 80+ providers". It bills inference at each provider's list price and takes its own fee when you buy credits. The fee is 5.5% on its Standard tier and 8% on Business. Both figures are from OpenRouter Pricing, read 19 November 2026.

The fee is a percentage of the purchase, so it grows with what you spend. On a $200 credit purchase it is $11 on Standard and $16 on Business. A small test project barely notices it. A team buying credits every month does.

That is a fee on top of tokens, not a different token price. Whether it matters depends on the second thing, which is the price of the model itself.

What Openference charges on top

Openference does not mark up tokens: credits and plan overage bill at the catalog rates in the tables below, and there is no per-request fee. Buying credits does carry a processing fee of 5% at checkout, and the full amount you buy is credited to your balance. On a $200 purchase that fee is $10, against $11 on OpenRouter's Standard tier.

The two fees are close, so the credit fee is not a reason to pick either gateway. The token rates, and whether you would rather pay a flat monthly price, are.

Why OpenRouter is not in the model tables

Because OpenRouter bills each provider's list price, the rate for a given model depends on which provider ends up serving it. We have no single OpenRouter number for a model, and we would rather leave it out than guess one. The tables below compare the gateways that publish one rate per model: Openference, Avian, R9S, AZET and Vercel AI Gateway, where they carry the model.

If you are comparing against OpenRouter in particular, open its page for the model you want, note the provider you would use, and put that rate next to the figures below. Then add the credit fee on top, which is 5.5% there on Standard and 5% at Openference.

The same model at different gateways

Each table prices one agent session on one model: 40 turns, an opening prompt of 12,000 tokens that grows by 1,500 tokens a turn, and 700 tokens written per turn. That is 1,650,000 input and 28,000 output tokens in all, without caching. The Claude Code guide explains where those numbers come from.

GLM-5.3: Openference is not the cheapest

GLM-5.3, one agent session, by gateway (no caching): 40 turns, a 12,000-token opening prompt that grows by 1,500 tokens a turn, and 700 tokens written per turn: 1,650,000 input and 28,000 output tokens in all.
GatewayPer 1M tokens, in / outSessionAgainst OpenferenceRates read
AZET$1.1319 / $3.5574$1.9719% less19 November 2026
Avian$1.19 / $3.74$2.0715% less19 November 2026
R9S$1.19 / $3.74$2.0715% less19 November 2026
Openference$1.40 / $4.40$2.43—live catalog
Vercel AI Gateway$1.40 / $4.40$2.43about the same19 November 2026

Each competitor’s rates are as that gateway publishes them on the date shown, linked from its name. Openference’s are read from the live catalog at the most expensive hour of the day. Rows are ordered by session cost, so a gateway that is cheaper than Openference for this model appears above it.

For GLM-5.3 the tally is 3 of 4, counting the other gateways that price the session more than a percent below Openference. If the lowest published token rate for GLM-5.3 is all you need, this table sends you elsewhere. Openference adds no markup to its catalog rate, but on this model the catalog rate is higher than most of what the others publish.

MiniMax M3: Openference is the cheapest

MiniMax M3, one agent session, by gateway (no caching): 40 turns, a 12,000-token opening prompt that grows by 1,500 tokens a turn, and 700 tokens written per turn: 1,650,000 input and 28,000 output tokens in all.
GatewayPer 1M tokens, in / outSessionAgainst OpenferenceRates read
Openference$0.03 / $0.12$0.053—live catalog
Avian$0.255 / $1.02$0.449750% more19 November 2026
R9S$0.29 / $1.15$0.511866% more19 November 2026

Each competitor’s rates are as that gateway publishes them on the date shown, linked from its name. Openference’s are read from the live catalog at the most expensive hour of the day. Rows are ordered by session cost, so a gateway that is cheaper than Openference for this model appears above it.

On MiniMax M3 the tally is 0 of 2: none of the other gateways that publish a rate price the session below Openference. The reason is a discount: the catalog takes 90% off its list price at every hour of the day. Without it, Openference would price this session above both of the others. Catalog rates can change, which is why this page is regenerated from them.

Same gateways, different answer for each model. Run the comparison for the model you actually use, not for an average.

What a plan changes

Per-token price is only one way to buy. Openference also sells monthly plans: Minion is $20 a month and Boss is $80. Each covers an allowance across rolling windows, and use beyond it bills as on-demand credits at the same catalog rates, plus the processing fee when you buy those credits.

A gateway that only sells credits has nothing to set against this, so the question stops being "which token price is lowest" and becomes "do I want a flat monthly cost". If your usage is light and irregular, credits alone may cost less than a plan. If you run agents every day, price it both ways. The pricing page lists what each plan's allowance covers.

When OpenRouter is the better choice

OpenRouter is the better choice when breadth matters. By its own description it fronts hundreds of models from many providers, and it lets you route between them. Openference lists 50 billed text models. That is a curated catalog, not a marketplace, so if the model you need is not in it, the price comparison above does not help you.

Open its page, read the current terms, and decide from the rate for your model. The comparison pages for the other gateways are OpenRouter, Vercel AI Gateway, Avian, R9S and AZET. Each lists the rates we read and the day we read them.

For what an OpenAI-compatible gateway does, see the LLM gateway page. For the GLM and Kimi catalogs, see GLM and Kimi.

More guides

All guides are on the guides index.

Model catalog · Pricing · Comparisons · Cost guides · Integration guides · FAQ
Guides: GLM coding plan alternative · Free AI API · LLM gateway · LLM proxy · Cheap AI API · OpenAI-compatible API
Model APIs: GLM API · DeepSeek API · Kimi API · Qwen API · MiniMax API