Documentation Index
Fetch the complete documentation index at: https://mintlify.com/simonw/LLM/llms.txt
Use this file to discover all available pages before exploring further.
LLM ships with a built-in plugin for OpenAI’s API covering chat models, reasoning models, completion models, and embedding models. gpt-4o-mini (aliased to 4o-mini) is the default model used when you do not specify one.
Configuration
All OpenAI models require an API key. Obtain one from the OpenAI API keys page, then configure LLM:
Paste your key at the prompt. LLM stores it securely and passes it to the API automatically. You can also set the OPENAI_API_KEY environment variable instead.
OpenAI Language Models
Run llm models for the full list. The OpenAI models built into LLM are:
GPT-4o Series
| Model ID | Aliases | Notes |
|---|
gpt-4o | 4o | Flagship multimodal model; images + PDFs |
chatgpt-4o-latest | chatgpt-4o | Always points to the latest ChatGPT-4o |
gpt-4o-mini | 4o-mini | Default model; least expensive |
gpt-4o-audio-preview | — | Accepts audio (MP3, WAV) |
gpt-4o-audio-preview-2024-12-17 | — | Dated snapshot |
gpt-4o-audio-preview-2024-10-01 | — | Dated snapshot |
gpt-4o-mini-audio-preview | — | Mini audio variant |
gpt-4o-mini-audio-preview-2024-12-17 | — | Dated snapshot |
GPT-4.1 Series
| Model ID | Aliases |
|---|
gpt-4.1 | 4.1 |
gpt-4.1-mini | 4.1-mini |
gpt-4.1-nano | 4.1-nano |
GPT-4 Series (Legacy)
| Model ID | Aliases |
|---|
gpt-4 | 4, gpt4 |
gpt-4-32k | 4-32k |
gpt-4-turbo | gpt-4-turbo-preview, 4-turbo, 4t |
gpt-4-turbo-2024-04-09 | — |
gpt-4-1106-preview | — |
gpt-4-0125-preview | — |
gpt-4.5-preview | gpt-4.5 |
gpt-4.5-preview-2025-02-27 | — |
GPT-3.5 Series (Legacy)
| Model ID | Aliases |
|---|
gpt-3.5-turbo | 3.5, chatgpt |
gpt-3.5-turbo-16k | chatgpt-16k, 3.5-16k |
o-Series Reasoning Models
These use the Responses API endpoint and support a reasoning_effort option (low, medium, high):
| Model ID | Aliases | Streaming |
|---|
o1 | — | ✗ |
o1-2024-12-17 | — | ✗ |
o1-preview | — | ✓ |
o1-mini | — | ✓ |
o3-mini | — | ✓ |
o3 | — | ✓ |
o4-mini | — | ✓ |
GPT-5 Series
| Model ID | Notes |
|---|
gpt-5 | Flagship |
gpt-5-mini | — |
gpt-5-nano | — |
gpt-5-2025-08-07 | Dated snapshot |
gpt-5-mini-2025-08-07 | Dated snapshot |
gpt-5-nano-2025-08-07 | Dated snapshot |
gpt-5.1, gpt-5.1-chat-latest | — |
gpt-5.2, gpt-5.2-chat-latest | — |
gpt-5.4, gpt-5.4-mini, gpt-5.4-nano | — |
gpt-5.5 | — |
Model Features
| Feature | Chat models | o-series (Responses) | Completion |
|---|
| Streaming | ✓ | Varies | ✓ |
| Async | ✓ | ✓ | ✗ |
| Image attachments | Selected models | Selected models | ✗ |
| PDF attachments | Selected models | Selected models | ✗ |
| Schemas | Selected models | ✓ | ✗ |
| Tools | Selected models | ✓ | ✗ |
Run llm models --options to see the exact set of features and accepted attachment types for each model.
gpt-4o accept images and PDFs via -a. Use llm models --options to see which types a specific model accepts.
Schemas — Models with schema support return JSON matching a provided JSON Schema. See the schemas documentation.
Options — Common options include temperature, max_tokens, top_p, frequency_penalty, presence_penalty, seed, and json_object. Reasoning models also expose reasoning_effort.
OpenAI Embedding Models
Run llm embed-models to list embedding models. LLM includes the following OpenAI embedding models:
| Model | Alias(es) | Vector size |
|---|
text-embedding-ada-002 | ada, ada-002 | 1536 |
text-embedding-3-small | 3-small | 1536 |
text-embedding-3-large | 3-large | 3072 |
text-embedding-3-small-512 | 3-small-512 | 512 |
text-embedding-3-large-256 | 3-large-256 | 256 |
text-embedding-3-large-1024 | 3-large-1024 | 1024 |
3-small model is the most cost-effective. The -512, -256, and -1024 variants are pre-truncated versions of the 3-small and 3-large models — smaller vectors reduce storage and query cost with a modest accuracy trade-off.
OpenAI Completion Models (Legacy)
gpt-3.5-turbo-instruct (aliases: 3.5-instruct, chatgpt-instruct) is a completion model designed to continue text rather than respond to chat-style messages.
llm -m gpt-3.5-turbo-instruct 'Reasons to tame a wild beaver:'
-o logprobs N option (unavailable on chat models), which stores the N most likely token log-probabilities in the database for each returned token.
Adding More OpenAI Models
OpenAI releases new model names regularly. While LLM ships updates to cover them, you can also add models immediately via a configuration file.
Find the config directory
dirname "$(llm logs path)"
~/Library/Application Support/io.datasette.llm.Create extra-openai-models.yaml
Create a file called extra-openai-models.yaml in that directory.
Add a model entry
- model_id: gpt-3.5-turbo-0613
model_name: gpt-3.5-turbo-0613
aliases: ["0613"]
model_id — the identifier stored in LLM’s logs and used on the command linemodel_name — the exact string passed to the OpenAI APIaliases — optional short names
Additional flags:| Key | Description |
|---|
completion: true | Use the completions endpoint instead of chat |
supports_schema: true | Model supports JSON schema output |
reasoning: true | Reasoning model (like o1, o3-mini) |
Verify the model is available
llm models
llm -m 0613 'What is the capital of France?'
o1-pro is not accessible through the Chat Completions API used by LLM’s default plugin. Install llm-openai-plugin to access it.