Supported models¶

The complete list of providers and models available today, grouped by what they're used for. Defaults are marked.

Current at time of writing

Models change often. This page is a snapshot — the live model registry in Kiwi is always the source of truth.

Chat & reasoning¶

Provider	Models
OpenAI	`gpt-4o`, `gpt-4o-mini`, `gpt-5`, `gpt-5-mini` (default), `gpt-5-nano`, `gpt-5.1`, `gpt-5.1-codex`, `gpt-5.2`, `gpt-5.2-codex`, `gpt-5.3-codex`, `gpt-5.4`, `gpt-5.4-nano`, `o3`, `o3-mini`, `o4-mini`
Google Gemini	`gemini-2.5-flash`, `gemini-2.5-pro`, `gemini-3-flash-preview`, `gemini-3.1-pro-preview`

Reasoning depth is configurable on models that support it.

Image generation¶

Provider	Models
Google	`nano-banana-2` (default), `nano-banana-2-lite`, `nano-banana-pro`
OpenAI	`gpt-image-1.5`, `gpt-image-2`
Adobe Firefly	`firefly-5`

Options include aspect ratio, size (1K/2K/4K, Google models) and quality (medium/high/auto, OpenAI models). Omitted settings use the effective defaults below.

Image editing¶

Provider	Models
Google	`nano-banana-2` (default), `nano-banana-2-lite`, `nano-banana-pro`
OpenAI	`gpt-image-1.5`
Adobe Firefly	`firefly-5`

Video generation¶

Provider	Models
Google Veo / Gemini Video	`gemini-omni-1.1-flash` (default), `veo-3.1-fast-generate-001`, `veo-3.1-generate-001`
Runway (Seedance variants)	`seedance2`, `seedance2_fast`, `seedance2_mini`

Options include aspect ratio (9:16, 16:9), duration (3–15s), resolution (360p/480p/720p/1080p/4K) and frame animation. Exact choices depend on the selected model.

Audio generation¶

Type	Provider	Models
Music	Google Lyria	`lyria-002`, `lyria-3-clip-preview` (default), `lyria-3-pro-preview`
Sound effects	ElevenLabs	default sound-effect model
Speech	ElevenLabs	`eleven_v3` (default), `eleven_multilingual_v2`, `eleven_flash_v2_5`, `eleven_turbo_v2_5`

Effective media defaults¶

When a setting is omitted, Kiwi applies these defaults for the selected model. Explicit requested settings take priority when the model supports them.

Operation	Model family	Effective defaults
Image generation	Google Nano Banana	`1:1`, `1K`.
Image generation	OpenAI	`1:1`, `quality=auto`.
Image generation	Firefly 5	`1:1` for text-only generation.
Video generation	Google Veo	`16:9`, 8 seconds, `1080p`.
Video generation	Seedance	`16:9`, 5 seconds, `480p`.
Video generation	Omni	`16:9`, 5 seconds, `720p` for an initial generation.
Sound effects	ElevenLabs	5 seconds.
Speech	ElevenLabs	Roger voice (`CwhRBWXzGAHq8TQ4Fs17`).

Conditioning can replace a shape default. Google and OpenAI image generation can derive shape from reference images when no ratio was requested; Firefly edits use their references; Omni follow-ups use their previous interaction; Visual Edit preserves the source image's framing.

Captioning¶

Capability	Providers
Image & video captioning	OpenAI, Google Gemini (default)

Provider notes¶

Anthropic (Claude) — coming soon as a new chat provider.
Azure OpenAI — not planned.
gpt-image-2 is available for generation only, not editing.

Related: Model · Generate media · Coming soon