Included models

Every account can use two hosted models without connecting a provider. OpenLLM pays the upstream provider, up to a monthly allowance that depends on your plan.

Both support chat, tool calls, streaming, reasoning, and image input. They run on the hosted gateway. A request sent to the daemon on your device is forwarded to the gateway for these models.

Monthly allowance

Allowances are counted in units, explained in the next section.

Each model has its own allowance. Both reset on the 1st of each month at 00:00 UTC. For what each plan includes, see Plans and limits.

How usage is counted

One unit is one fresh input token. Other tokens are weighted by their list price relative to input:

For example, a DeepSeek request with 10,000 fresh input tokens, 50,000 cached input tokens, and 2,000 output tokens uses 10,000 + (50,000 × 0.02) + (2,000 × 4) = 19,000 units.

Only successful requests count. Usage on the included models never counts toward the Free plan's monthly token cap. If a Free account hits that cap on other models, though, every request is blocked until it resets, including requests to the included models.

Context window

OpenLLM treats both models as having a 250,000-token input window, even though the upstream models accept more. Under Hop to larger context, the default, a request that's clearly larger than that skips these models when another model comes after them in the chain. See Context overflow.

When the allowance is spent

Before each request, OpenLLM removes included models whose allowance is spent. Routing falls through to the next model in the chain, as if the included model weren't there.

If no other model is left, for example when you call openllm/qwen/qwen3.7-plus directly, the request fails with 403. The gateway currently reports this with the type subscription_requires_daemon and a message about subscription providers. See Errors.

To keep working after the allowance runs out, add a model from one of your own providers to the chain. See Fallback chains.

Where they appear in your chains

On a new account, the built-in chains start with the included models:

• ultra and plus start with openllm/qwen/qwen3.7-plus, then openllm/deepseek/deepseek-v4.1-flash.
• lite starts with openllm/deepseek/deepseek-v4.1-flash.

Models from providers you connect follow them. Once you edit a chain, it's yours and OpenLLM doesn't reorder it. See Aliases.

See what's left

The Plan tile in the app's sidebar shows Included models: one bar for both models, and a line per model with the units left and the monthly limit. A model's numbers turn red when its allowance is used up.

Who serves them

• Qwen 3.7 Plus runs on Alibaba DashScope, on OpenLLM's DashScope account.
• DeepSeek V4.1 Flash runs through OpenRouter, on OpenLLM's OpenRouter account. OpenLLM limits OpenRouter to endpoints priced at or below DeepSeek's list price.

Your prompts and responses pass through OpenLLM's servers to these providers. For what is stored and for how long, see Privacy.

Related

Last updated on