Models — FreeModel

Modality routes

These are the part a plain model list does not have. Instead of tracking which image or speech model is available this month, you send auto/image and the choice happens server-side.

What you give up is knowing which model answered. A route promises a modality, not a model. If you need to reproduce a result exactly, or name a model for compliance reasons, send a specific id instead. Routes are for when you want the right answer more than you want to know where it came from.

Route	What the name says it selects for
`auto/image`	an image-generation model
`auto/video`	a video model
`auto/tts`	a text-to-speech model
`auto/asr`	a speech-to-text model
`auto/embed`	an embedding model
`auto/rerank`	a rerank model

The exact model each route resolves to is decided at request time, so it can change as providers add, retire, or reprice models.

What the models cover

Modality counts across the 199 listed models, read from the same endpoint:

Modality	Models
Chat	153
Image	18
Video	12
Text-to-speech	6
Speech-to-text	4
Embedding	4
Rerank	2

Chat is 77% of the list. The other 23% is where an image, a video or a voice lands, and each of those modalities has its own route.

The endpoint states a modality for every entry, so these counts cover the whole list rather than a subset of it.

How to see the list yourself

The endpoint is public. No key, no account, no rate limit worth mentioning.

# every model, as JSON
curl -s https://freemodel.online/v1/models

# just the modality routes
curl -s https://freemodel.online/v1/models | grep -o '"auto/[^"]*"'

# how many there are, without trusting this page
curl -s https://freemodel.online/v1/models | grep -o '"id"' | wc -l

Which ones are free

Free quota comes and goes as providers run promotions. See which models have some right now →

Finding a model for a job

The list is long enough that browsing it is the wrong approach. Query it instead, in this order.

Step 1: Start from the job, not from a name. Decide first whether the request carries tools, images, or a document larger than a context window; that decides the filters before quality is even a question.
Step 2: Query the list and filter locally. curl -s https://freemodel.online/v1/models returns every entry with the metadata the gateway attaches — an id, a modality, and a context length on some entries — so the filtering happens on your machine.
Step 3: If you have no strong preference, send a route and skip the choice entirely. Six of the entries are modality routes (3% of the list) and each one encodes exactly this decision.[1]
Step 4: If you do have a preference, pin the model id. A pinned id is reproducible and auditable; a route is neither, by design.

What counts as an entry here

The endpoint returns 199 entries, and they are not all the same kind of object: 153 are chat models, and the rest span image, video, speech, embeddings and rerank. It is the modality, not a capability flag, that the list states for every entry. 77% of the 199 listed models are chat models and 9% generate images[1], which is why the split is worth reading before you filter.

The shape of the list is worth knowing before you filter it. Chat is the bulk of it, at 77%, and the other 23% is split across the remaining modalities. A request that names a modality outside chat is choosing from a much smaller pool, and the modality is the tightest filter in the list.

Read the modality as a filter, not a ranking. It says what kind of object an entry is. It says nothing about how well it handles a request.

What the list does not tell you

It does not rank quality. The modality describes what an entry is, not how good it is. Treating the list as a scoreboard is the most common way to misread this page.
It does not carry prices. /v1/models is the callable list. Rates live in the separate catalogue endpoint, and the two are deliberately different sizes.
It does not report free quota. The response has no free-quota field, because free capacity changes by the hour. It is checked when the request is made, and the free page explains what that means.
It does not report uptime. An entry being present means it is callable now, not that it will answer in a minute. That is what failover is for.

Models FAQ

Do I need a key to read the model list?

No. GET /v1/models is public and unauthenticated, which is why this page can tell you to count the entries yourself. Sending an actual request to one of those models does need a key.

Why do two entries look like the same model?

Usually because they are different versions, or the same family served by different hosts. The id is the identity — count by id if you need a total, and treat similar names as separate entries until their modality says otherwise.

Is a modality route a model?

It is an entry in the list, but not a model you can pin. Send auto/image and the router resolves it to a concrete model at request time. Six of the entries are routes, and they are the reason the first question to ask is whether you need a reproducible model id or a good answer.

What happens when a model disappears from the list?

It stops being callable, and it stops appearing — the list is live rather than a catalogue page. Code that pins that id starts failing; code that sends a route moves to another candidate and records the skip.

Are the modality counts stable?

No. They move as providers add, retire and reprice models, which is why each figure here is tied to the endpoint rather than to a date. If a count matters to a decision, count it from /v1/models on the day you decide.

Sources

Every third-party figure quoted on this site was read from the vendor documentation linked below, on the date stated. None of these links are affiliate links and none of these vendors paid to be listed. If a number here disagrees with the page it came from, the page is right — tell us and it gets corrected.

Where the percentages come from

Every share quoted on this site is counted from the live model list, not estimated. Divide the two counts in the last column, round to the nearest whole percent, and you get the number in the first.

Share	Of what	Count
77%	are chat models	153 of 199 listed entries
9%	generate images	18 of 199 listed entries
23%	are not chat models	46 of 199 listed entries
3%	are modality routes	6 of 199 listed entries
17%	carry a context length	34 of 199 listed entries

[1] Counted on 29 September 2026 from GET https://freemodel.online/v1/models, which needs no key and no account. Every row uses all 199 listed entries as the denominator, because the only thing the endpoint states for every entry is its modality. It no longer publishes per-model capability flags such as tool calling or reasoning, so this site quotes no shares for them. Fetch the list and recount whenever you want to check the arithmetic.

200+ models, and 6 ways to stop choosing one