openllmsopenllms

Open LLMs

Every open model, one click away.

Start for freeExplore models

Models

  • Qwen-Coder-3

    Chat

    Chat, tool calls and agents.

    Context
    66K tokens
    Parameters
    80B (3B active)
    Licence
    Apache-2.0
    Output
    Text

    $0.30 / 1M tokens

  • GLM-4.5-Air

    Chat

    Chat, tool calls and agents.

    Context
    33K tokens
    Parameters
    106B (12B active)
    Licence
    MIT
    Output
    Text

    $1.50 / 1M tokens

  • Wan2.1-VACE Video

    Alibaba (Wan)

    Text/identity-to-video generation (Wan2.1-VACE).

    Parameters
    14B
    Licence
    apache-2.0
    Output
    Video

    $50.00 / 1M tokens

  • Whisper Large v3

    OpenAI

    Turns speech into text.

    Parameters
    809M
    Licence
    MIT
    Output
    Text

    $0.004 / minute

  • Qwen-2.5

    Alibaba (Qwen)

    Chat, tool calls and agents.

    Context
    33K tokens
    Parameters
    500M
    Licence
    Apache-2.0
    Output
    Text

    $0.15 / 1M tokens

OpenAI-compatible

Two lines change. Nothing else.

Point the OpenAI SDK you already use at Open LLMs and pick a model from the table above. Same request shape, same response shape, same streaming.

Base URL
https://api.openai.com/v1https://api.openllms.co/v1
Model
gpt-4oglm-4-5-air-awq
Read the API reference
curl https://api.openllms.co/v1/chat/completions \
  -H "Authorization: Bearer $OPENLLMS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-4-5-air-awq",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

About Open LLMs

The short version

What is Open LLMs?

Open LLMs is an OpenAI-compatible API for open-weight models. You point an OpenAI client at our base URL, use an Open LLMs key, and pay per request from a prepaid balance in US dollars — no subscription, no seat fee, no minimum commitment.

The request and response shapes are the ones you already use, so adopting it is a base URL and a model id rather than a rewrite. If it turns out not to suit you, leaving is the same two lines in reverse.

The honest case

Why open models?

Because the weights are open, there is no licence fee priced into a request. The cost is the machine that ran it, which is why an open model of a given capability is usually cheaper to serve than a closed one of the same capability.

That is also the limit of the argument. On a task where a frontier closed model is genuinely better, the cheaper request is not the better one, and no amount of price advantage fixes an answer you cannot use. Pick per workload — what the API is used for sets out which models suit which job, and says what each is not for.

No surprises

How am I charged?

Per request, in the unit that matches what the model does: chat and vision per million tokens, speech-to-text per minute of audio, video per second of output. A model billed by the minute is never quoted to you in tokens.

You top up a wallet and requests are metered against it. There is nothing to cancel and nothing renews. Every current rate is read from the same catalogue that serves the requests, so the page and the bill cannot disagree.

Before you start

What do I need?

An API key and a wallet with something in it. The first call to a model that has been idle can answer 503 while it starts — wait for the Retry-After header and retry, and the second call succeeds.

The API reference covers every endpoint and what it accepts. Every error the API can return lists each status with its code, why it happened and what to do — worth a bookmark before you need it rather than after.