FreeModel vs OpenRouter

The short answer

Side by side

Where OpenRouter is the better choice

This section is the part of this page worth trusting. If it reads like a pitch, the rest of the page is worth less.

1. The model you need is probably there

Their list carries 460 public models; ours carries 199. If you need a specific model and it is not in our list, theirs almost certainly has it. Check both lists before you commit: GET /v1/models works without a key on either service.

2. You want to know exactly which model you are on

Both services draw free capacity from upstream supply; neither runs its own free inference. The difference is who chooses. OpenRouter makes you name the :free model explicitly, so you always know which one you are using and what its quota is. FreeModel's routing makes that choice for you: simpler to use, but you are not the one picking.

3. You are plugging into an existing tool

Most third-party tools, tutorials and Stack Overflow answers assume OpenRouter. If you are following someone else's setup guide, sticking with it means fewer surprises.

4. You want a published fee model

5.5% on credit purchases, $0.80 minimum, documented on their FAQ. You can model the cost before you commit.

Where FreeModel is the better choice

1. You run Claude Code, Cursor, or anything expecting Anthropic format

Point the base URL at us directly. No translation layer in between, and no second endpoint to maintain.

2. You do not want to think about :free

OpenRouter's free models are a whitelist you opt into on every request. Ours are the default: a request goes to free capacity if there is any, and falls through when there is not.

3. You are orchestrating agents, not just calling a model

Four agent clients ship preset (Claude Code, Codex CLI, Hermes and OpenClaw) with task dispatch and an approval step for consequential actions. That layer is the product; the gateway is underneath it.

4. You want to try before you pay

No credit purchase required to start.

Switching in either direction

From OpenRouter to FreeModel

base_url    https://freemodel.online/v1              (OpenAI format)
            https://freemodel.online/api/gateway     (Anthropic format)

model names drop the :free suffix — routing is free-first by default,
with no per-model opt-in.

list models first:  GET /v1/models   (readable without a key)

From FreeModel to OpenRouter

base_url    https://openrouter.ai/api/v1

model names add :free explicitly, or the request will not use a free
model — the opposite of how FreeModel behaves.

What is behind each model count

A count on its own does not tell you whether a list fits your work. Both numbers here are public and readable without a key, so the useful question is what each one is made of.

The 199 ids on our public list break down as 77% chat models, 9% image generation and 23% everything else[1] — video, speech synthesis, speech recognition, embeddings and rerank, which is 153, 18 and 46 entries. Six of the entries are modality routes: auto/image, auto/video, auto/tts, auto/asr, auto/embed and auto/rerank. Each of those carries a candidates count, so you can read how deep its fallback is before you call it.[1] OpenRouter publishes 460 models on its own public list; we have not broken that list down the same way, so the two totals are not strictly comparable line for line.

Read the shares as a description of the inventory, not a quality claim. The 23% of ids that are not chat models[1] cannot answer a coding question, and a list padded with models you will never call is worth less than a shorter one you can read in a single sitting.

Switching without guessing

Whichever direction you move, three steps cover it. The trap is treating the base URL as the whole job when it is the first line of it.

1. Step 1: move the base URL and the key. Toward us that is https://freemodel.online/v1 for OpenAI-format clients and https://freemodel.online/api/gateway for Anthropic-format ones, with a key from the console. Toward OpenRouter it is https://openrouter.ai/api/v1.
2. Step 2: fix the model names. Coming to us, drop the :free suffix and read GET /v1/models to see what each id is called here. Going the other way, the suffix becomes mandatory — without it a request will not use a free model, which is the opposite of the default here.
3. Step 3: enforce anything that must not move. Routing decides which model answers unless you name one. If a prompt, an eval or a benchmark depends on a specific model, pin the id; if it must never cost money, send fm-v1-lite, the free tier any account can call with no subscription, rather than relying on the free-first preference.

FreeModel vs OpenRouter FAQ