---
title: "Models"
description: "Select models, compare outputs, and configure defaults."
canonical_url: "https://ai-cli.dev/docs/models"
---
# Models

## Default models

Each command type has a default model used when `-m` is not specified:

| Type          | Default                  | Env override                 |
| ------------- | ------------------------ | ---------------------------- |
| text          | `openai/gpt-5.5`         | `AI_CLI_TEXT_MODEL`          |
| image         | `openai/gpt-image-2`     | `AI_CLI_IMAGE_MODEL`         |
| video         | `bytedance/seedance-2.0` | `AI_CLI_VIDEO_MODEL`         |
| speech        | `openai/tts-1`           | `AI_CLI_SPEECH_MODEL`        |
| transcription | `openai/whisper-1`       | `AI_CLI_TRANSCRIPTION_MODEL` |

The `-m` flag always takes priority over environment variables.

## Specifying a model

Use `creator/model-name` format:

```bash
ai text -m "anthropic/claude-sonnet-4" "hello"
ai image -m "bfl/flux-2-pro" "a sunset"
ai audio speak -m "openai/tts-1" "hello"
```

Short names (without the creator prefix) are resolved against models fetched from the gateway:

```bash
ai text -m gpt-5.5 "hello"
ai image -m flux-2-pro "a sunset"
ai audio speak -m tts-1 "hello"
```

Model IDs must contain printable ASCII characters without spaces. This applies to both `--model` values and the `AI_CLI_*_MODEL` environment variables. Invalid values are rejected before a request is sent.

## Multi-model comparison

Generate with multiple models by comma-separating `-m`:

```bash
ai image "a sunset" -m "openai/gpt-image-1,xai/grok-imagine-image,bfl/flux-2-pro"
```

Combine with `-n` to generate multiple per model:

```bash
ai image "a sunset" -n 2 -m "openai/gpt-image-1,bfl/flux-2-pro"
```

This produces 4 images total (2 per model), saved with generated filename stems and numeric suffixes, such as `resp_abc-1.png` through `resp_def-4.png`.

## Listing models

```bash
ai models                          # all models
ai models --type image             # image models only
ai models --type audio             # speech and transcription models
ai models --creator google         # Google models only
ai models --json                   # JSON with descriptions
```

In human mode, models are grouped by creator and display short names. Use `--json` for machine-readable output with full IDs and descriptions.

## Model info

Pass a model ID (or short name) to `ai models` to see its details — context window, max output tokens, pricing (input, output, cache read/write, web search), release date and per-provider latency, throughput and uptime:

```bash
ai models claude-opus-4.6                    # detail card
ai models anthropic/claude-opus-4.6 --json   # machine-readable, includes raw pricing and endpoints
```

```
Claude Opus 4.6  anthropic/claude-opus-4.6
Released 2026-02-05 · tool-use · reasoning · vision · web-search

  Context      1M
  Max output   128K
  Input        $5/M
  Output       $25/M
  Cache read   $0.5/M
  Cache write  $6.25/M
  Web search   $10/K + input costs

Providers
  provider   context  latency  throughput  uptime
  anthropic  1M       1.4s     49tps       99.9%
  bedrock    1M       1.4s     56tps       99.9%
```

Latency (time to first token) and throughput are p50 over the last hour; uptime covers the last day.

## Providers

All models are accessed through [Vercel AI Gateway](https://vercel.com/docs/ai-gateway):

- **OpenAI**: GPT-5.5, GPT-4.1, GPT Image 2, TTS, Whisper

- **Anthropic**: Claude Sonnet 4

- **Google**: Gemini 2.5 Pro, Imagen 4, Veo 3

- **xAI**: Grok 3, Grok Imagine

- **Meta**: Llama 4 Maverick

- **Black Forest Labs**: FLUX 2 Pro, FLUX Kontext

- **ByteDance**: Seedance 2.0, Seedream 5.0

- **Kling AI**: Kling v3.0

- **Alibaba**: Wan v2.5, Wan v2.6

- **Recraft**: Recraft v2, v3, v4

- **Prodia**: FLUX Fast Schnell