OpenAI

When using OpenAi models, llm-exe will make POST requests to https://api.openai.com/v1/chat/completions. All models are supported if you pass openai.chat.v1 as the first argument, and then specify a model in the options.

Basic Usage

OpenAi Chat

const llm = useLlm("openai.chat.v1", {
  model: "gpt-4o", // specify a model
});

OpenAi Chat By Model

const llm = useLlm("openai.gpt-4o", {
  // other options,
  // no model needed, using gpt-4o
});

INFO

You can use the following models using this shorthand:

• openai.gpt-6
• openai.gpt-6-astra
• openai.gpt-5.6
• openai.gpt-5.6-terra
• openai.gpt-5.6-luna
• openai.gpt-5.5
• openai.gpt-5.4
• openai.gpt-5.4-mini
• openai.gpt-5.2
• openai.gpt-4.1
• openai.gpt-4.1-mini
• openai.gpt-4o
• openai.gpt-4o-mini
• openai.gpt-4.1-nanodeprecated
  Model "openai.gpt-4.1-nano" is deprecated and will shut down on 2026-10-23. Migrate to openai.gpt-5.6-luna.
• openai.gpt-4deprecated
  Model "openai.gpt-4" is deprecated and will shut down on 2026-10-23. Migrate to openai.gpt-4o.
• openai.gpt-5-minideprecated
  Model "openai.gpt-5-mini" is deprecated and will shut down on 2026-12-11. Migrate to "openai.gpt-5.6-terra".
• openai.gpt-5-nanodeprecated
  Model "openai.gpt-5-nano" is deprecated and will shut down on 2026-12-11. Migrate to "openai.gpt-5.6-luna".
• openai.o3deprecated
  Model "openai.o3" is deprecated and will shut down on 2026-12-11. Migrate to "openai.gpt-5.6".
• openai.o4-minideprecated
  Shorthand "openai.o4-mini" is deprecated and may be removed in a future release.

Deprecated shorthands still resolve and will continue to work for now, but the underlying provider may stop accepting them at any time. Migrate to a current shorthand when you can.

Authentication

To authenticate, you need to provide an OpenAi API Key. You can provide the API key various ways, depending on your use case.

1. Pass in as execute options using openAiApiKey
2. Pass in as setup options using openAiApiKey
3. Use a default key by setting an environment variable of OPENAI_API_KEY

Generally you pass the LLM instance off to an LLM Executor and call that. However, it is possible to interact with the LLM object directly, if you wanted.

// call the LLM directly with a prompt
await llm.call(prompt);

OpenAi-Specific Options

In addition to the generic options, the following options are OpenAi-specific and can be passed in when creating a llm function.

Option	Type	Default	Description
model	string	—	The model to use. Must be specified when using openai.chat.v1. Can be any valid chat model. See OpenAI Docs
openAiApiKey	string	undefined	API key for OpenAi. See authentication
topP	number	undefined	Maps to top_p. See OpenAI Docs
stopSequences	array	undefined	Maps to stop. See OpenAI Docs
frequencyPenalty	number	undefined	Maps to frequency_penalty. See OpenAI Docs
logitBias	object	undefined	Maps to logit_bias. See OpenAI Docs
useJson	boolean	undefined	When true, sets response_format to json_object
effort	string	undefined	Maps to reasoning_effort. Valid values: "none", "minimal", "low", "medium", "high". Only sent for OpenAI reasoning models — any model whose name starts with gpt-5, gpt-6, o3, or o4; on any other model, or for a value outside that list, it is silently dropped. "none" is accepted only by gpt-5.1 through gpt-5.6; it is rejected with a 400 by the original gpt-5, gpt-5-mini, and gpt-5-nano (floor "minimal"), and by gpt-6, o3, and o4-mini (floor "low").

See OpenAI API Reference for details on these parameters.

Reasoning models (gpt-5*, gpt-6*, o3*, o4*) rewrite maxTokens, temperature, and topP

OpenAI's reasoning models reject the legacy Chat Completions parameters, so llm-exe adjusts the request for any model whose name starts with gpt-5, gpt-6, o3, or o4:

• maxTokens is sent as max_completion_tokens instead of max_tokens. Reasoning models return a 400 on max_tokens (Use 'max_completion_tokens' instead). Non-reasoning models (gpt-4o, gpt-4.1, ...) still receive max_tokens.
• temperature and topP are dropped unless effort is "none". Reasoning models only accept the default sampling values while reasoning is active; gpt-5.5, gpt-5.6, gpt-6, and the o-series reject them even when reasoning_effort is omitted. Pass effort: "none" on gpt-5.1 through gpt-5.6 to keep them. There is no way to keep them on gpt-6, o3, or o4-mini: those models reject non-default sampling values and also reject "none", so llm-exe always drops them there.

Both rules are specific to the OpenAI provider. The xAI and DeepSeek providers, which share the same mapper, forward these parameters unchanged.