Skip to model catalog

Model catalog

30 executable

Every model, every parameter

This page renders the live catalog — the same models and JSON Schemas returned by GET /api/v1/models. Pass any model's id and a params object to POST /api/v1/generations. Media inputs accept either a Workspace Asset reference ({"asset_id": …}) or an HTTPS URL ({"url": …}).

KREA 2

krea-2 image preview

genaro · text_to_image

KREA 2 text-to-image generation with optional LoRAs.

Parameters

Parameter Details
prompt required

string

Image generation prompt.

aspect_ratio

string enum

1:1 4:3 3:2 16:9 21:9 3:4 2:3 9:16 4:5 5:4

Default "4:5"

loras

object

A LoRA stack: an optional preset, at most one character LoRA, then style LoRAs in order. Each entry names exactly one of lora_id, lora_version_id, curated_id. The model's `lora` block lists the kinds, slots, strength range, curated LoRAs and presets.

Default null

resolution

string enum

1k 2k

Default "2k"

seed

integer | null

nullable

Default null

Price · USD per generation

Parameters Price for this example
2k $0.09
1k $0.09

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with resolution loras

Minimal request

POST /api/v1/generations
{
  "model": "krea-2",
  "params": {
    "prompt": "<prompt>"
  }
}

Required parameters only — quote first with POST /generations/quote.

H3

minimax-h3 video preview

minimax · text_to_video · image_to_video

Self-hosted MiniMax H3 joint video and audio generation.

Parameters

Parameter Details
prompt required

string

Video generation prompt.

aspect_ratio

string enum

16:9 9:16 4:3 3:4 1:1 21:9

Default "9:16"

duration

integer

Video length in seconds.

5 – 14

Default 5

end_image

media input

Optional final frame.

{"asset_id": …} or {"url": "https://…"} · nullable

Default null

image

media input

Optional first frame.

{"asset_id": …} or {"url": "https://…"} · nullable

Default null

improve_prompt

boolean

Rewrite the prompt with MiniMax's official H3 prompt-writing skill before generation. Off by default; leave off when the prompt is already in H3's structured format.

Default false

resolution

string enum

768p is H3's native canvas; 480p is the faster, cheaper draft canvas.

480p 768p

Default "768p"

Price · USD per generation

Parameters Price for this example
5s, 768p $0.5093
5s, 480p $0.3176
10s, 768p $0.8686
14s, 768p $1.2872

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with duration resolution

Minimal request

POST /api/v1/generations
{
  "model": "minimax-h3",
  "params": {
    "prompt": "<prompt>"
  }
}

Required parameters only — quote first with POST /generations/quote.

SeeDream v5 Pro

seedream-v5-pro image preview

bytedance · text_to_image

ByteDance flagship v5 text-to-image model.

Parameters

Parameter Details
prompt required

string

aspect_ratio

string enum

1:1 1:2 2:1 1:3 3:1 2:3 3:2 3:4 4:3 4:5 5:4 9:16 16:9 9:21 21:9

Default "1:1"

output_format

string enum

jpeg png

Default "png"

resolution

string enum

1k 2k

Default "2k"

Price · USD per generation

Parameters Price for this example
2k $0.085
1k $0.042

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with resolution

Minimal request

POST /api/v1/generations
{
  "model": "seedream-v5-pro",
  "params": {
    "prompt": "<prompt>"
  }
}

Required parameters only — quote first with POST /generations/quote.

Motion Control 3

motion-control-3 video preview

genaro · video_to_video

Self-hosted character animation: a character image driven by the motion of a reference video, as long as that video.

Parameters

Parameter Details
image required

media input

The character to animate: a Workspace Asset or an HTTPS image URL.

{"asset_id": …} or {"url": "https://…"}

video required

media input

The driving video whose motion the character follows: a Workspace Asset with a known duration. The output is as long as this video, and priced per second of it.

{"asset_id": …} — Workspace Asset only

negative_prompt

string | null

Optional negative prompt. Blank by default.

nullable

Default null

prompt

string | null

Optional positive prompt. Blank by default.

nullable

Default null

Price · $0.10 USD per second of driving video

Parameters Price for this example
10s driving video $1.00 ($0.10 per second)

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with video

Minimal request

POST /api/v1/generations
{
  "model": "motion-control-3",
  "params": {
    "image": {
      "asset_id": "<asset-uuid>"
    },
    "video": {
      "asset_id": "<asset-uuid>"
    }
  }
}

Required parameters only — quote first with POST /generations/quote.

Face Merge

face-merge image preview

genaro · image_to_image

Blend two to four faces into an ordered set of frames, each at a known mix of the inputs, plus an optional MP4: a face morph, or a new face that merges several people.

Parameters

Parameter Details
images required

array of media inputs

Two to four face photographs, in order: face A, B, C, D. Each must show exactly one face; every face is aligned to the same headshot framing before blending.

2–4 items · each entry: {"asset_id": …} or {"url": "https://…"}

endpoint_noise

number

How much a frame of one face alone is re-synthesized (0.2 to 0.95, at most `noise`; left out, 0.4 or `noise` if that is lower). Lower keeps that face closer to its photo. Frames in between ramp from this up to `noise` as the mix evens out.

0.2 – 0.95

Default 0.4

fps

integer

Frames per second of the MP4 (1 to 60).

1 – 60

Default 12

frames

integer

Two faces: how many evenly spaced frames from all of face A to all of face B (2 to 100). Ignored for three or four faces and when `weights` is set.

2 – 100

Default 33

grid_divisions

integer

Three or four faces: the grid's step is 1 / grid_divisions (4 gives steps of 0.25). Three faces make a triangle of every mix in those steps (15 frames at 4). Four faces make a square with one face per corner, clockwise from the top left (A top left, B top right, C bottom right, D bottom left), and a frame at every point of a (grid_divisions + 1) by (grid_divisions + 1) grid over it (25 frames at 4, at most 9 divisions). Ignored for two faces and when `weights` is set.

1 – 10

Default 4

lora

string enum

The realism LoRA every frame is denoised with: `realistic_snapshot` (natural skin texture and soft light instead of a polished, plastic look) or `none` (the base KREA 2 Turbo model).

realistic_snapshot none

Default "realistic_snapshot"

lora_strength

number

How strongly the LoRA applies (0 to 2). Ignored when `lora` is `none`. Higher looks more photographic.

0 – 2

Default 1.5

mp4

boolean | null

Also render the frames, in order, as an MP4 (needs at least two frames). Defaults to on for two faces with two or more frames, and off otherwise.

nullable

Default null

noise

number

How much a full blend (no face above half the mix) is re-synthesized (0.3 to 0.95). Lower keeps more of the inputs; higher cleans up ghosting between faces. 0.8 works best with the default LoRA and 8 steps.

0.3 – 0.95

Default 0.8

prompt

string

Describes the shot the frames are cleaned up toward. Describe the subjects for the best results (for example "a young woman with glamorous makeup"); the default is a flattering beauty portrait of an unnamed person.

Default "a close-up beauty portrait photograph of an attractive person with glowing smooth skin and polished styling, looking at the camera, head and shoulders centered, soft flattering studio lighting, sharp focus"

prompt_strength

number

How much each frame is pulled toward the prompt (0 to 2): true classifier-free guidance against an empty prompt. 1 is the prompt as written; below 1 follows it less and keeps more of the faces, above 1 follows it harder. Any value other than 1 runs two passes per step and doubles the denoising part of the price.

0 – 2

Default 1.0

seed

integer | null

Random seed. Blank = random; set to reproduce.

nullable

Default null

size

string enum

Frame size in pixels, width x height.

768x1024 1024x1024

Default "768x1024"

steps

integer

Denoising steps per frame, through the tail of the schedule that `noise` enters (4 to 8). More steps give finer detail and cost more per frame.

4 – 8

Default 8

weights

array | null

Your own frames instead: one weight vector per frame, in order, each with one non-negative weight per image summing to 1 (1 to 100 frames).

nullable

Default null

Price · USD per generation

Parameters Price for this example
two faces, 33 frames $0.4287
two faces, 33 frames at 1024x1024 $0.4703
two faces, 33 frames, 4 steps $0.3825
two faces, 33 frames, prompt strength 1.5 $0.5211
three faces, steps of 0.25 (15 frames) $0.3585
four faces, steps of 0.25 (25 frames) $0.3975

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with images frames grid_divisions weights size steps prompt_strength

Minimal request

POST /api/v1/generations
{
  "model": "face-merge",
  "params": {
    "images": [
      {
        "asset_id": "<asset-uuid>"
      }
    ]
  }
}

Required parameters only — quote first with POST /generations/quote.

H3 Ref

minimax-h3-ref video preview

minimax · image_to_video

Self-hosted MiniMax H3 reference-to-video: up to nine ordered reference images guide a generated clip with native stereo sound.

Parameters

Parameter Details
prompt required

string

Video generation prompt.

images required

array of media inputs

One to nine ordered reference images. Order is meaningful — the prompt may address references by position.

1–9 items · each entry: {"asset_id": …} or {"url": "https://…"}

aspect_ratio

string enum

Canvas aspect; references do not bind the output geometry.

16:9 9:16 4:3 3:4 1:1 21:9

Default "9:16"

audio_lock

boolean

Hard-lock exactly one explicit audio reference into H3's target-audio latent throughout denoising. The candidate uses that original reference audio instead of H3's resynthesized soundtrack. Requires exactly one entry in reference_audios: an audio file or a video whose soundtrack to use.

Default false

duration

integer

Video length in seconds (5-14), generated at a fixed 24 fps and snapped UP to the model's `17n+5` frame grid. 14 is the maximum: the model caps at 360 frames, so 15 s would snap past that limit.

5 – 14

Default 5

improve_prompt

boolean

Rewrite the prompt with MiniMax's official H3 prompt-writing skill before generation. Off by default; leave off when the prompt is already in H3's structured format.

Default false

reference_audios

array of media inputs

Up to three ordered reference audio clips (2–15.5s each). The clip is generated against this audio, so a sung or spoken track drives the performance instead of the model inventing its own. An entry may also be a video, in which case its soundtrack is used.

up to 3 items · each entry: {"asset_id": …} or {"url": "https://…"}

Default []

reference_videos

array of media inputs

Up to three ordered reference videos (2–15.5s each). Drives motion and camera (MiniMax's V2V motion transfer) while identity comes from the image references.

up to 3 items · each entry: {"asset_id": …} or {"url": "https://…"}

Default []

resolution

string enum

768p is H3's native canvas. 480p is the fast preview: half-resolution canvas and a 10-step sampler, several times faster and cheaper to run. Expect soft detail and rough audio; re-render at 768p for final quality.

480p 768p

Default "768p"

Price · USD per generation

Parameters Price for this example
5s, 768p, one reference image $0.50
5s, 480p, one reference image $0.3289
10s, 768p, one reference image $1.00
14s, 768p, one reference image $1.40

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with duration resolution images reference_videos reference_audios

Minimal request

POST /api/v1/generations
{
  "model": "minimax-h3-ref",
  "params": {
    "prompt": "<prompt>",
    "images": [
      {
        "asset_id": "<asset-uuid>"
      }
    ]
  }
}

Required parameters only — quote first with POST /generations/quote.

Nano Banana Pro

nano-banana-pro image available

google · text_to_image · image_to_image

Google's Nano Banana Pro: text-to-image, or an edit of up to 14 source images.

Parameters

Parameter Details
prompt required

string

Image prompt, or the edit instruction.

aspect_ratio

string | null

Output aspect ratio: 1:1 when omitted, 4:5 for an edit.

nullable

Default null

images

array of media inputs

Optional ordered source images, each an asset_id or an https url, at most 14. Given any, the model edits the first.

0–14 items · each entry: {"asset_id": …} or {"url": "https://…"}

Default []

output_format

string enum

png jpeg

Default "png"

resolution

string enum

Output resolution tier.

1k 2k 4k

Default "2k"

Price · USD per generation

Parameters Price for this example
2k $0.13
1k $0.13
4k $0.22
2k, one source image $0.13
4k, one source image $0.22

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with resolution

Minimal request

POST /api/v1/generations
{
  "model": "nano-banana-pro",
  "params": {
    "prompt": "<prompt>"
  }
}

Required parameters only — quote first with POST /generations/quote.

ElevenLabs Text to Speech

elevenlabs-text-to-speech audio preview

elevenlabs · text_to_speech

Natural speech from text in a chosen voice, optionally with per-character timestamps. Priced per character of text, by speech model.

Parameters

Parameter Details
text required

string

The text to speak, up to 5000 characters.

voice_id required

string enum

ElevenLabs voice ID, one of the available voices (`enum`; `x-voices` names them).

CwhRBWXzGAHq8TQ4Fs17 EXAVITQu4vr4xnSDxMaL FGY2WhTYpPnrIDTdsKH5 IKne3meq5aSn9XLyUdCD JBFqnCBsd6RMkjVDRZzb N2lVS1w4EtoT3dr4eOWO SAz9YHcvj6GT2YYXdXww SOYHLrjzK2X1ezoPC6cr TX3LPaxmHKxFdv7VOQHJ Xb7hH8MSUJpSbSDYk0k2 XrExE9yKIg1WjnnlVkGX bIHbv24MWmeRgasZH58o cgSgspJ2msm6clMCkdW9 cjVigY5qzO86Huf0OWal hpp4J3VqNfWAUOO0d1Us iP95p4xoKVk53GoZ742B nPczCjzI2devNBz1zQrb onwK4e9ZLuTAKqWW03F9 pFZP5JQG7iQjIQuC4Bku pNInz6obpgDQGcFmaJgB pqHfZKP75CvOlQylNhV4

model_id

string enum

Speech model. Flash and Turbo are fastest and half the price; Multilingual v2, v3 and v4 are the most expressive.

eleven_multilingual_v2 eleven_v3 eleven_v4 eleven_flash_v2_5 eleven_turbo_v2_5

Default "eleven_multilingual_v2"

output_format

string enum

Audio encoding of the output Asset (MP3 at the given sample rate and bitrate).

mp3_44100_192 mp3_44100_128 mp3_44100_64 mp3_22050_32

Default "mp3_44100_128"

similarity_boost

number

How closely to hold to the voice, 0 to 1.

0 – 1

speed

number

Speaking speed, 0.7 through 1.2 (1.0 is normal).

0.7 – 1.2

stability

number

Voice stability, 0 (more variable) to 1 (more even).

0 – 1

style

number

Style exaggeration, 0 to 1.

0 – 1

use_speaker_boost

boolean

Boost similarity to the original speaker.

with_timestamps

boolean

When true the finished run's `output.data.alignment` holds per-character timings (`characters`, `character_start_times_seconds`, `character_end_times_seconds`, in seconds from the start of the audio), and `output.data.normalized_alignment` when ElevenLabs returns one. The audio Asset and the price are the same.

Default false

Price · USD per generation

Parameters Price for this example
1,000 characters, Multilingual v2 $0.092
1,000 characters, v3 $0.092
1,000 characters, v4 $0.092
1,000 characters, Flash v2.5 $0.046
5,000 characters, Multilingual v2 $0.46

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with text model_id

Minimal request

POST /api/v1/generations
{
  "model": "elevenlabs-text-to-speech",
  "params": {
    "text": "<text>",
    "voice_id": "CwhRBWXzGAHq8TQ4Fs17"
  }
}

Required parameters only — quote first with POST /generations/quote.

ElevenLabs Sound Effects

elevenlabs-sound-effects audio preview

elevenlabs · text_to_sound_effect

Sound effects from a text description. Priced per second of duration.

Parameters

Parameter Details
prompt required

string

Describe the sound, up to 450 characters.

duration

integer

Length of the sound in seconds, 1 through 30.

1 – 30

Default 5

Price · USD per generation

Parameters Price for this example
5s $0.0115
10s $0.023
30s $0.069

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with duration

Minimal request

POST /api/v1/generations
{
  "model": "elevenlabs-sound-effects",
  "params": {
    "prompt": "<prompt>"
  }
}

Required parameters only — quote first with POST /generations/quote.

ElevenLabs Music

elevenlabs-music audio preview

elevenlabs · text_to_music

Original music from a text description. Priced per second of duration.

Parameters

Parameter Details
prompt required

string

Describe the music: genre, mood, instruments, structure.

duration

integer

Length of the track in seconds, 3 through 300.

3 – 300

Default 30

instrumental

boolean

Guarantee a track without vocals.

Default false

Price · USD per generation

Parameters Price for this example
30s $0.0863
60s $0.1725
300s $0.8625

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with duration

Minimal request

POST /api/v1/generations
{
  "model": "elevenlabs-music",
  "params": {
    "prompt": "<prompt>"
  }
}

Required parameters only — quote first with POST /generations/quote.

ElevenLabs Voice Changer

elevenlabs-voice-changer audio preview

elevenlabs · speech_to_speech

Re-voices recorded speech, keeping the performance. Priced per second of source audio.

Parameters

Parameter Details
audio required

media input

Source speech, an audio Asset or an HTTPS URL, up to 300 seconds.

{"asset_id": …} or {"url": "https://…"}

voice_id required

string enum

ElevenLabs voice ID to speak in, one of the available voices (`enum`; `x-voices` names them).

CwhRBWXzGAHq8TQ4Fs17 EXAVITQu4vr4xnSDxMaL FGY2WhTYpPnrIDTdsKH5 IKne3meq5aSn9XLyUdCD JBFqnCBsd6RMkjVDRZzb N2lVS1w4EtoT3dr4eOWO SAz9YHcvj6GT2YYXdXww SOYHLrjzK2X1ezoPC6cr TX3LPaxmHKxFdv7VOQHJ Xb7hH8MSUJpSbSDYk0k2 XrExE9yKIg1WjnnlVkGX bIHbv24MWmeRgasZH58o cgSgspJ2msm6clMCkdW9 cjVigY5qzO86Huf0OWal hpp4J3VqNfWAUOO0d1Us iP95p4xoKVk53GoZ742B nPczCjzI2devNBz1zQrb onwK4e9ZLuTAKqWW03F9 pFZP5JQG7iQjIQuC4Bku pNInz6obpgDQGcFmaJgB pqHfZKP75CvOlQylNhV4

model_id

string enum

Speech-to-speech model.

eleven_multilingual_sts_v2 eleven_english_sts_v2

Default "eleven_multilingual_sts_v2"

output_format

string enum

Audio encoding of the output Asset (MP3 at the given sample rate and bitrate).

mp3_44100_192 mp3_44100_128 mp3_44100_64 mp3_22050_32

Default "mp3_44100_128"

remove_background_noise

boolean

Remove background noise from the source before converting.

similarity_boost

number

How closely to hold to the voice, 0 to 1.

0 – 1

stability

number

Voice stability, 0 to 1.

0 – 1

style

number

Style exaggeration, 0 to 1.

0 – 1

Price · $0.0023 USD per second of source audio

Parameters Price for this example
10s of source audio $0.023 ($0.0023 per second)

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with audio

Minimal request

POST /api/v1/generations
{
  "model": "elevenlabs-voice-changer",
  "params": {
    "audio": {
      "asset_id": "<asset-uuid>"
    },
    "voice_id": "CwhRBWXzGAHq8TQ4Fs17"
  }
}

Required parameters only — quote first with POST /generations/quote.

ElevenLabs Audio Isolation

elevenlabs-audio-isolation audio preview

elevenlabs · audio_isolation

Strips background noise and music, leaving the voice. Priced per second of source audio.

Parameters

Parameter Details
audio required

media input

Source audio, an audio Asset or an HTTPS URL, up to 300 seconds.

{"asset_id": …} or {"url": "https://…"}

Price · $0.0023 USD per second of source audio

Parameters Price for this example
10s of source audio $0.023 ($0.0023 per second)

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with audio

Minimal request

POST /api/v1/generations
{
  "model": "elevenlabs-audio-isolation",
  "params": {
    "audio": {
      "asset_id": "<asset-uuid>"
    }
  }
}

Required parameters only — quote first with POST /generations/quote.

Seedance 2.5

seedance-v2-5 video preview

bytedance · text_to_video · image_to_video

ByteDance Seedance 2.5 video with native audio, up to 1080p: from text, a first frame, a first and last frame, or reference images, videos and audio.

Parameters

Parameter Details
aspect_ratio

string | null

Output aspect ratio, 16:9 when omitted. Not with a start frame: the video keeps its shape.

nullable

Default null

draft

boolean

A cheap, fast 480p preview of the shot (draft mode). Leave resolution out or send 480p. Works with a start frame, reference images, videos and audio. Render the approved take at 1080p with `from_draft`, within 7 days. One output per request.

Default false

duration

integer | string

Video length in seconds.

4 – 30

Default 5

end_image

media input

Optional end frame; needs `image`.

{"asset_id": …} or {"url": "https://…"} · nullable

Default null

from_draft

string | null

The id of a succeeded draft Generation (`draft: true`) in this Workspace: renders that take at 1080p on the provider that made it, reusing its prompt, inputs, duration, aspect ratio and audio. Send it alone: every other param comes from the draft. Priced as the same request at 1080p.

nullable

Default null

generate_audio

boolean

Generate native audio synchronized with the video.

Default true

image

media input

Optional start frame. The video keeps its shape.

{"asset_id": …} or {"url": "https://…"} · nullable

Default null

images

array of media inputs

Optional reference images guiding style, characters or composition. Not with a start frame.

up to 9 items · each entry: {"asset_id": …} or {"url": "https://…"}

Default []

prompt

string

reference_audios

array of media inputs

Optional reference audio clips, at most 30 seconds in total. Asset clips must have a known duration and are checked at admission; a URL clip is judged by the provider. Not with a start frame.

up to 9 items · each entry: {"asset_id": …} or {"url": "https://…"}

Default []

reference_videos

array of media inputs

Optional reference videos, at most 30 seconds in total. Each must be a Workspace Asset with a known duration: reference seconds are billed. Not with a start frame.

up to 9 items · each entry: {"asset_id": …} or {"url": "https://…"}

Default []

resolution

string enum

480p 720p 1080p

Default "720p"

Price · USD per generation

Parameters Price for this example
5s, 720p $1.75
5s, 480p $0.85
5s, 1080p $4.35
15s, 720p $5.25
5s, 720p, one 10s reference video $3.225
5s, 1080p, one 10s reference video $7.95

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with resolution duration reference_videos

Minimal request

POST /api/v1/generations
{
  "model": "seedance-v2-5",
  "params": {
    "prompt": "<prompt>"
  }
}

Required parameters only — quote first with POST /generations/quote.

Wan 3.0 T2V

wan-3-0-t2v video preview

alibaba · text_to_video

Alibaba Wan 3.0 text-to-video, with native audio.

Parameters

Parameter Details
prompt required

string

aspect_ratio

string enum

16:9 9:16 1:1 4:3 3:4

Default "16:9"

duration

integer | string

Video length in seconds.

2 – 30

Default 5

resolution

string enum

480p 720p 1080p

Default "720p"

Price · USD per generation

Parameters Price for this example
5s, 720p $0.45
5s, 480p $0.225
5s, 1080p $0.90
15s, 720p $1.35

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with resolution duration

Minimal request

POST /api/v1/generations
{
  "model": "wan-3-0-t2v",
  "params": {
    "prompt": "<prompt>"
  }
}

Required parameters only — quote first with POST /generations/quote.

Wan 3.0 I2V

wan-3-0-i2v video preview

alibaba · image_to_video

Alibaba Wan 3.0 image-to-video, with native audio.

Parameters

Parameter Details
prompt required

string

image required

media input

Required start frame.

{"asset_id": …} or {"url": "https://…"}

aspect_ratio

string | null

Output aspect ratio; adapts to the input image when omitted.

nullable

Default null

duration

integer | string

Video length in seconds.

2 – 30

Default 5

end_image

media input

Optional end frame.

{"asset_id": …} or {"url": "https://…"} · nullable

Default null

resolution

string enum

480p 720p 1080p

Default "720p"

Price · USD per generation

Parameters Price for this example
5s, 720p $0.45
5s, 480p $0.225
5s, 1080p $0.90
15s, 720p $1.35

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with resolution duration

Minimal request

POST /api/v1/generations
{
  "model": "wan-3-0-i2v",
  "params": {
    "prompt": "<prompt>",
    "image": {
      "asset_id": "<asset-uuid>"
    }
  }
}

Required parameters only — quote first with POST /generations/quote.

Wan 3.0 Reference-to-Video

wan-3-0-reference-to-video video preview

alibaba · text_to_video

Alibaba Wan 3.0 reference-to-video: a prompt plus reference images, videos or audio. Reference videos are billed as input seconds on top of the output, with native audio.

Parameters

Parameter Details
prompt required

string

aspect_ratio

string enum

16:9 9:16 1:1 4:3 3:4

Default "16:9"

duration

integer | string

Video length in seconds.

2 – 30

Default 5

images

array of media inputs

Up to ten reference images guiding subject, style or composition.

up to 10 items · each entry: {"asset_id": …} or {"url": "https://…"}

Default []

reference_audios

array of media inputs

Up to five reference audio clips (15 seconds combined). Asset clips must have a known duration and are checked at admission; a URL clip is judged by the provider.

up to 5 items · each entry: {"asset_id": …} or {"url": "https://…"}

Default []

reference_videos

array of media inputs

Up to five reference videos (15 seconds combined). Each must be a Workspace Asset with a known duration: reference seconds are billed, and they plus `duration` may not pass 30 seconds.

up to 5 items · each entry: {"asset_id": …} or {"url": "https://…"}

Default []

resolution

string enum

480p 720p 1080p

Default "720p"

Price · USD per generation

Parameters Price for this example
5s, 720p $0.45
5s, 480p $0.225
5s, 1080p $0.90
15s, 720p $1.35
5s, 720p, one 5s reference video $0.90

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with resolution duration reference_videos

Minimal request

POST /api/v1/generations
{
  "model": "wan-3-0-reference-to-video",
  "params": {
    "prompt": "<prompt>"
  }
}

Required parameters only — quote first with POST /generations/quote.

GPT Image 2.5 Flare

gpt-image-2-5-flare image available

openai · text_to_image · image_to_image

OpenAI's GPT Image 2.5 Flare, the fast tier: text-to-image, or an edit of up to 14 source images.

Parameters

Parameter Details
prompt required

string

Image prompt, or the edit instruction.

aspect_ratio

string | null

Output aspect ratio: 1:1 when omitted, or the first source image's shape for an edit.

nullable

Default null

images

array of media inputs

Optional ordered source images, each an asset_id or an https url, at most 14. Given any, the model edits the first.

0–14 items · each entry: {"asset_id": …} or {"url": "https://…"}

Default []

quality

string enum

Image quality: high, xhigh or max; higher qualities cost more.

high xhigh max

Default "high"

resolution

string enum

Output resolution tier.

1k 2k 4k

Default "2k"

Price · USD per generation

Parameters Price for this example
high quality, 2k $0.135
high quality, 1k $0.08
max quality, 4k $0.86
high quality, 2k, one source image $0.145
max quality, 4k, 14 source images $0.87

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with quality resolution images

Minimal request

POST /api/v1/generations
{
  "model": "gpt-image-2-5-flare",
  "params": {
    "prompt": "<prompt>"
  }
}

Required parameters only — quote first with POST /generations/quote.

GPT Image 2.5 Sunburst

gpt-image-2-5-sunburst image available

openai · text_to_image

OpenAI's GPT Image 2.5 text-to-image model, the slower, higher-fidelity tier, with quality- and resolution-aware pricing.

Parameters

Parameter Details
prompt required

string

aspect_ratio

string enum

1:1 1:2 2:1 1:3 3:1 2:3 3:2 3:4 4:3 4:5 5:4 9:16 16:9 9:21 21:9

Default "1:1"

output_format

string enum

png jpeg webp

Default "png"

quality

string enum

Image quality; higher qualities cost more.

low medium high xhigh max

Default "medium"

resolution

string enum

Output resolution tier.

1k 2k 4k

Default "2k"

Price · USD per generation

Parameters Price for this example
medium quality, 2k $0.055
low quality, 1k $0.035
max quality, 4k $0.86

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with quality resolution

Minimal request

POST /api/v1/generations
{
  "model": "gpt-image-2-5-sunburst",
  "params": {
    "prompt": "<prompt>"
  }
}

Required parameters only — quote first with POST /generations/quote.

GPT Image 2.5 Sunburst Edit

gpt-image-2-5-sunburst-edit image available

openai · image_to_image

OpenAI's GPT Image 2.5 image editor, the slower, higher-fidelity tier; priced per edit plus a small charge per extra source image.

Parameters

Parameter Details
prompt required

string

Edit instruction.

images required

array of media inputs

One to 16 ordered source images, each an asset_id or an https url. The first is the one edited.

1–16 items · each entry: {"asset_id": …} or {"url": "https://…"}

aspect_ratio

string | null

Output aspect ratio. Omit to follow the first source image.

nullable

Default null

output_format

string enum

png jpeg webp

Default "png"

quality

string enum

Image quality; higher qualities cost more.

low medium high xhigh max

Default "medium"

resolution

string enum

Output resolution tier.

1k 2k 4k

Default "2k"

Price · USD per generation

Parameters Price for this example
medium quality, 2k $0.055
low quality, 1k $0.035
max quality, 4k $0.87
max quality, 4k, 16 source images $0.87

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with quality resolution images

Minimal request

POST /api/v1/generations
{
  "model": "gpt-image-2-5-sunburst-edit",
  "params": {
    "prompt": "<prompt>",
    "images": [
      {
        "asset_id": "<asset-uuid>"
      }
    ]
  }
}

Required parameters only — quote first with POST /generations/quote.

GPT Image 2.6 Flare

gpt-image-2-6-flare image preview

openai · text_to_image · image_to_image

OpenAI's GPT Image 2.6 Flare, the fast tier: text-to-image, or an edit of up to 14 source images.

Parameters

Parameter Details
prompt required

string

Image prompt, or the edit instruction.

aspect_ratio

string | null

Output aspect ratio: 1:1 when omitted, or the first source image's shape for an edit.

nullable

Default null

complexity

string enum

How much the model plans the layout before drawing. Does not change the price.

low medium high

Default "high"

images

array of media inputs

Optional ordered source images, each an asset_id or an https url, at most 14. Given any, the model edits the first.

0–14 items · each entry: {"asset_id": …} or {"url": "https://…"}

Default []

quality

string enum

Image quality; higher qualities cost more.

low medium high xhigh max

Default "medium"

resolution

string enum

Output resolution tier.

1k 2k 4k

Default "2k"

Price · USD per generation

Parameters Price for this example
medium quality, 2k $0.14
low quality, 1k $0.04
max quality, 4k $1.12
medium quality, 2k, one source image $0.143
max quality, 4k, 14 source images $1.162

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with quality resolution images

Minimal request

POST /api/v1/generations
{
  "model": "gpt-image-2-6-flare",
  "params": {
    "prompt": "<prompt>"
  }
}

Required parameters only — quote first with POST /generations/quote.

GPT Image 2.6 Sunburst

gpt-image-2-6-sunburst image preview

openai · text_to_image

OpenAI's GPT Image 2.6 text-to-image model, the slower, higher-fidelity tier, with quality- and resolution-aware pricing.

Parameters

Parameter Details
prompt required

string

aspect_ratio

string enum

1:1 1:2 2:1 1:3 3:1 2:3 3:2 3:4 4:3 4:5 5:4 9:16 16:9 9:21 21:9

Default "1:1"

complexity

string enum

How much the model plans the layout before drawing. Does not change the price.

low medium high

Default "high"

output_format

string enum

png jpeg webp

Default "png"

quality

string enum

Image quality; higher qualities cost more.

low medium high xhigh max

Default "medium"

resolution

string enum

Output resolution tier.

1k 2k 4k

Default "2k"

Price · USD per generation

Parameters Price for this example
medium quality, 2k $0.14
low quality, 1k $0.04
max quality, 4k $1.12

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with quality resolution

Minimal request

POST /api/v1/generations
{
  "model": "gpt-image-2-6-sunburst",
  "params": {
    "prompt": "<prompt>"
  }
}

Required parameters only — quote first with POST /generations/quote.

GPT Image 2.6 Sunburst Edit

gpt-image-2-6-sunburst-edit image preview

openai · image_to_image

OpenAI's GPT Image 2.6 image editor, the slower, higher-fidelity tier; priced per edit plus a small charge per source image.

Parameters

Parameter Details
prompt required

string

Edit instruction.

images required

array of media inputs

One to 16 ordered source images, each an asset_id or an https url. The first is the one edited.

1–16 items · each entry: {"asset_id": …} or {"url": "https://…"}

aspect_ratio

string | null

Output aspect ratio. Omit to follow the first source image.

nullable

Default null

complexity

string enum

How much the model plans the layout before drawing. Does not change the price.

low medium high

Default "high"

output_format

string enum

png jpeg webp

Default "png"

quality

string enum

Image quality; higher qualities cost more.

low medium high xhigh max

Default "medium"

resolution

string enum

Output resolution tier.

1k 2k 4k

Default "2k"

Price · USD per generation

Parameters Price for this example
medium quality, 2k $0.143
low quality, 1k $0.043
max quality, 4k $1.123
max quality, 4k, 16 source images $1.2535

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with quality resolution images

Minimal request

POST /api/v1/generations
{
  "model": "gpt-image-2-6-sunburst-edit",
  "params": {
    "prompt": "<prompt>",
    "images": [
      {
        "asset_id": "<asset-uuid>"
      }
    ]
  }
}

Required parameters only — quote first with POST /generations/quote.

Nano Banana 2.1

nano-banana-2.1 image preview

google · text_to_image · image_to_image

Google's Nano Banana 2.1: text-to-image, or an edit of up to ten source images.

Parameters

Parameter Details
prompt required

string

Image prompt, or the edit instruction.

aspect_ratio

string | null

Output aspect ratio: 1:1 when omitted, 4:5 for an edit.

nullable

Default null

images

array of media inputs

Optional ordered source images, each an asset_id or an https url, at most 10. Given any, the model edits the first.

0–10 items · each entry: {"asset_id": …} or {"url": "https://…"}

Default []

output_format

string enum

png jpeg

Default "png"

resolution

string enum

Output resolution tier.

1k 2k 4k

Default "2k"

Price · USD per generation

Parameters Price for this example
2k $0.055
1k $0.035
4k $0.12
2k, one source image $0.055
2k, ten source images $0.055

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with resolution images

Minimal request

POST /api/v1/generations
{
  "model": "nano-banana-2.1",
  "params": {
    "prompt": "<prompt>"
  }
}

Required parameters only — quote first with POST /generations/quote.

SeeDream v5 Pro Edit

seedream-v5-pro-edit image available

bytedance · image_to_image

ByteDance's SeeDream v5 Pro image editor, with resolution-aware pricing plus a charge per extra source image.

Parameters

Parameter Details
prompt required

string

Edit instruction.

images required

array of media inputs

One to 10 ordered source images, each an asset_id or an https url. The first is the one edited.

1–10 items · each entry: {"asset_id": …} or {"url": "https://…"}

aspect_ratio

string | null

Output aspect ratio. Omit to follow the first source image.

nullable

Default null

output_format

string | null

Output file format; omit for the provider default.

nullable

Default null

resolution

string enum

Output resolution tier.

1k 2k

Default "2k"

Price · USD per generation

Parameters Price for this example
2k, one source image $0.085
1k, one source image $0.042
2k, ten source images $0.1346

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with resolution images

Minimal request

POST /api/v1/generations
{
  "model": "seedream-v5-pro-edit",
  "params": {
    "prompt": "<prompt>",
    "images": [
      {
        "asset_id": "<asset-uuid>"
      }
    ]
  }
}

Required parameters only — quote first with POST /generations/quote.

SeeDream v5.5

seedream-v5-5 image preview

bytedance · text_to_image

ByteDance's SeeDream v5.5 text-to-image model, up to 2K, priced per image by resolution.

Parameters

Parameter Details
prompt required

string

aspect_ratio

string enum

1:1 4:3 3:4 16:9 9:16 3:2 2:3 21:9

Default "1:1"

background

string | null

transparent gives a PNG with an alpha channel (needs png output); omit to let the prompt decide.

nullable

Default null

output_format

string enum

png jpeg

Default "png"

resolution

string enum

Output resolution tier; 1k and 1.5k cost the same.

1k 1.5k 2k

Default "1k"

Price · USD per generation

Parameters Price for this example
1k $0.0195
1.5k $0.0195
2k $0.034

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with resolution

Minimal request

POST /api/v1/generations
{
  "model": "seedream-v5-5",
  "params": {
    "prompt": "<prompt>"
  }
}

Required parameters only — quote first with POST /generations/quote.

SeeDream v5.5 Edit

seedream-v5-5-edit image preview

bytedance · image_to_image

ByteDance's SeeDream v5.5 image editor, up to 2K, priced per image by resolution plus a charge per source image.

Parameters

Parameter Details
prompt required

string

Edit instruction.

images required

array of media inputs

One to 10 ordered source images, each an asset_id or an https url. The first is the one edited; each is billed.

1–10 items · each entry: {"asset_id": …} or {"url": "https://…"}

aspect_ratio

string | null

Output aspect ratio. Omit to follow the first source image.

nullable

Default null

background

string | null

transparent gives a PNG with an alpha channel (needs png output); omit to let the prompt decide.

nullable

Default null

output_format

string enum

png jpeg

Default "png"

resolution

string enum

Output resolution tier; 1k and 1.5k cost the same.

1k 1.5k 2k

Default "1k"

Price · USD per generation

Parameters Price for this example
1k, one source image $0.0273
2k, one source image $0.0418
2k, ten source images $0.112

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with resolution images

Minimal request

POST /api/v1/generations
{
  "model": "seedream-v5-5-edit",
  "params": {
    "prompt": "<prompt>",
    "images": [
      {
        "asset_id": "<asset-uuid>"
      }
    ]
  }
}

Required parameters only — quote first with POST /generations/quote.

Enhancor Skin Fix V4

enhancor-skin-fix-v4 image preview

enhancor · image_to_image

Enhancor's skin retoucher for one photo, no prompt: 4K Base, 4K Fast or 8K Ultra, priced per run by mode.

Parameters

Parameter Details
image required

media input

The photo to fix: one JPEG or PNG, an asset_id or an https url.

{"asset_id": …} or {"url": "https://…"}

enhancement_strength

string | null

The look of the fix; fast and ultra only. Omit for realistic.

nullable

Default null

fix_lighting

boolean

Also fixes the lighting; fast and ultra only, with the realistic strength.

Default false

freckle_intensity

string | null

How many freckles; fast and ultra only, with the freckle strength. Omit for 50.

nullable

Default null

mode

string enum

base: 4K. fast: a faster 4K. ultra: 8K. Each mode has its own price.

base fast ultra

Default "base"

Price · USD per generation

Parameters Price for this example
4K Base $0.1622
4K Fast $0.2277
8K Ultra $1.3598

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with mode

Minimal request

POST /api/v1/generations
{
  "model": "enhancor-skin-fix-v4",
  "params": {
    "image": {
      "asset_id": "<asset-uuid>"
    }
  }
}

Required parameters only — quote first with POST /generations/quote.

Enhancor Ultrafine

enhancor-ultrafine image preview

enhancor · image_to_image

Enhancor's upscaler for one photo, no prompt: Full multiplies its size up to 16x, Fast makes a 2432 px photo, and Skin fix keeps faces real. Priced per output megapixel.

Parameters

Parameter Details
image required

media input

The photo to upscale: an asset_id of a JPEG, PNG or WebP of at most 24 MB and 4096 x 4096 px, at least 256 px on its shorter side.

{"asset_id": …} — Workspace Asset only

mode

string enum

full: multiplies the photo's size by scale. fast: a photo 2432 px on its longer side, quickly; best on photos under 1000 px.

full fast

Default "full"

scale

string | null

How many times bigger each side gets; full only. 8 and 16 need skin_fix off. Omit for 2. The output may be at most 48 megapixels.

nullable

Default null

skin_fix

boolean

Keeps faces looking like real skin, at a higher rate per megapixel. Full goes up to 4x with it.

Default false

Price · USD per generation

Parameters Price for this example
Full 2x, 1024 x 1024 photo $0.2655
Full 2x, 1024 x 1024 photo, skin fix $0.3377
Full 4x, 1024 x 1024 photo $1.062
Full 4x, 1024 x 1024 photo, skin fix $1.3506
Fast, 1024 x 1024 photo $0.3744
Fast, 1024 x 1024 photo, skin fix $0.4762

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with image mode skin_fix scale

Minimal request

POST /api/v1/generations
{
  "model": "enhancor-ultrafine",
  "params": {
    "image": {
      "asset_id": "<asset-uuid>"
    }
  }
}

Required parameters only — quote first with POST /generations/quote.

Hailuo 3 T2V

hailuo-3-t2v video preview

minimax · text_to_video

MiniMax Hailuo 3 text-to-video with a synchronized soundtrack, up to 2k.

Parameters

Parameter Details
prompt required

string

The scene and its motion.

aspect_ratio

string enum

16:9 9:16 4:3 3:4 1:1 21:9

Default "16:9"

duration

integer | string

Video length in seconds.

4 – 15

Default 5

resolution

string enum

480p 768p 2k

Default "768p"

Price · USD per generation

Parameters Price for this example
5s, 768p $0.45
5s, 480p $0.225
5s, 2k $0.65
15s, 2k $1.95

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with resolution duration

Minimal request

POST /api/v1/generations
{
  "model": "hailuo-3-t2v",
  "params": {
    "prompt": "<prompt>"
  }
}

Required parameters only — quote first with POST /generations/quote.

Hailuo 3 I2V

hailuo-3-i2v video preview

minimax · image_to_video

MiniMax Hailuo 3 image-to-video with a synchronized soundtrack, up to 2k.

Parameters

Parameter Details
prompt required

string

The scene and its motion.

image required

media input

Required first frame.

{"asset_id": …} or {"url": "https://…"}

aspect_ratio

string | null

Output aspect ratio; adapts to the first frame when omitted.

nullable

Default null

duration

integer | string

Video length in seconds.

4 – 15

Default 5

end_image

media input

Optional last frame.

{"asset_id": …} or {"url": "https://…"} · nullable

Default null

resolution

string enum

480p 768p 2k

Default "768p"

Price · USD per generation

Parameters Price for this example
5s, 768p $0.45
5s, 480p $0.225
5s, 2k $0.65
15s, 2k $1.95

A run is charged at most its quoted price, and usually exactly. It can settle below the quote at completion, never above: when a length held at its maximum is measured, or the provider that served the run reports a lower cost, the difference is refunded. The quote for your request: POST /generations/quote.

Price varies with resolution duration

Minimal request

POST /api/v1/generations
{
  "model": "hailuo-3-i2v",
  "params": {
    "prompt": "<prompt>",
    "image": {
      "asset_id": "<asset-uuid>"
    }
  }
}

Required parameters only — quote first with POST /generations/quote.

What "preview" means

Every model on this page is executable: true: it has a registered provider route and a price, and POST /generations accepts it. availability: "preview" promises only that — no quality, latency or capacity guarantee yet, and its parameters may still change. availability: "available" marks a model that has also had a successful live generation recorded. Failed generations are refunded in full either way.

GET /models?include_hidden=true also returns catalog entries that cannot run yet (executable: false, availability: "inventory") and models held out of this list until they are proven. Submitting a generation against an executable: false entry returns 422 model_unavailable and charges nothing. A held-back model that reports executable: true still accepts generations.