Skip to content
Layer Docs

list_base_models

List the base AI models available for generation in a workspace.

Sourced from the model registry and the workspace's feature config — not the Style table —
so every enabled base model is listed. Results are ranked by the requested `sort`, which
defaults to Layer's curated recommendation order.

To choose a model for a task, pass `filter.use_case` (and any required capabilities) and
take the first result. When the task has craft knowledge attached to it — pixel art,
storyboards, seamless textures, and similar — load the matching `model-guidance` skill for
prompt and workflow technique before generating.

IMPORTANT: Call tool `get_instructions` first for general platform guidance.

Parameters

ParameterTypeDefaultDescription
filterModelQueryFilter—Query filters
fulltextstring—Search model name/description
modalityenum—Filter by modality: image, video, three_d, audio, world, text. `text` is served only by tools that can place a text-generation step and is rejected as invalid input everywhere else.
One of: image, video, three_d, audio, world, text
capabilitiesCapabilityFilter—Filter by model capabilities
supports_reference_imagesboolean—Set true to require at least one supported guidance file type — any type at all, so this alone does not guarantee any particular one. Omit (don't set false) to skip this filter
image_editingboolean—Set true to require image-editing support — models that take a source image plus reference images to reproduce subjects from. Use this when generating new art that must match provided references. Omit to skip
style_referenceboolean—Set true to require style-reference (IP-Adapter `prompt` type) support — legacy style-only transfer that does not reproduce subjects. Omit (don't set false) to skip this filter
character_poseboolean—Set true to require character pose reference images. Omit to skip
structureboolean—Set true to require depth/structure reference images. Omit to skip
outlineboolean—Set true to require edge/outline reference images. Omit to skip
line_artboolean—Set true to require line art reference images. Omit to skip
squareboolean—Set true to require 1:1 square output; omit to skip
portrait_9_16boolean—Set true to require 9:16 portrait output; omit to skip
landscape_16_9boolean—Set true to require 16:9 landscape output; omit to skip
multiple_aspect_ratiosboolean—Set true to require more than one aspect ratio; omit to skip
inpaintingboolean—Set true to require inpainting support. Omit (don't set false) to skip this filter
outpaintingboolean—Set true to require outpainting support; omit to skip
tileabilityboolean—Set true to require seamless tileable textures; omit to skip
tile_gridboolean—Set true to require a model that draws game tiles (isometric or hex) onto a fixed grid (`tile_grid` on a generation); omit to skip
negative_promptboolean—Set true to require negative prompt support; omit to skip
generate_audioboolean—Set true to require audio generation with video; omit to skip
lipsyncboolean—Set true to require lip-sync in video; omit to skip
video_effectsboolean—Set true to require video effects; omit to skip
text_to_3dboolean—Set true to require text-to-3D; omit to skip
image_to_3dboolean—Set true to require image-to-3D; omit to skip
texturesboolean—Set true to require textured 3D models; omit to skip
text_to_speechboolean—Set true to require text-to-speech; omit to skip
musicboolean—Set true to require a music model rather than a sound-effect one; omit to skip
instrumentalboolean—Set true to require a flag that guarantees no vocals. Prose asking for an instrumental is only a suggestion, so require this whenever the track must not sing. Omit to skip
loopboolean—Set true to require seam-free looping the provider composes rather than merely describes; omit to skip
composition_planboolean—Set true to require a model that takes the track as a list of sections, each with its own length, styles and exclusions. Required whenever the arc has to be specified rather than narrated in the prompt. Omit to skip
fixed_durationboolean—Set true to require a model that accepts a requested clip length, excluding those that choose their own and ignore duration_seconds. Required for anything that has to land on an exact length, such as a loop. Omit to skip
sound_effectboolean—Set true to require sound effect generation; omit to skip
use_caseenum—Filter to models that declare this use case, e.g. text_to_image, image_editing, image_to_video, upscaling. This is the primary way to narrow a task-shaped request.
One of: text_to_image, image_editing, multi_angle, inpainting, reframing, split_into_layers, text_to_video, image_to_video, video_editing, video_extend, insert_video, video_subtitle, video_lipsync, motion_transfer, video_object_swap, video_character_swap, relight_video, image_to_video_lipsync, text_to_3d, image_to_3d, text_to_world, image_to_world, remesh, retexture, uv_unwrap, animate, rig, segment, image_segment, text_to_speech, sound_effects, music, upscaling, video_interpolation, denoise_image, denoise_video, denoise_audio, voice_change, audio_separation, audio_isolation, deblur_video, restore_image, sdr_to_hdr_video, background_removal, vectorization, virtual_try_on, restyle, pixelate, colorize, change_expression, erase_object, edit_text, dub_video, extend_audio, generate_pbr_materials, create_skybox, complete_draft
providerstring—Filter by provider slug, e.g. 'bfl', 'google'. Case-insensitive.
featured_onlyboolean—Set true to return only models in the curated featured set. Omit to include everything.
max_pricenumber>0—Exclude models whose average Creative Units per generation exceed this value.
released_afterstring (date)—Exclude models publicly released on or before this date (ISO 8601, e.g. 2026-01-01).
guidance_file_typesenum[]—Keep only models that accept every listed guidance input, e.g. `first_frame` and `reference_video` together. `get_base_model` gives each slot's file and length limits.
sortenumrecommendedResult ordering. 'recommended' (default) uses Layer's curated ranking for the requested use case or modality, falling back to quality score then popularity for uncurated models; 'elo' orders by Artificial Analysis quality score; 'popularity' by usage on Layer; 'newest' by public release date.
One of: recommended, elo, popularity, newest
paginationPaginationInput—Parameters driving result pagination
limitinteger1–10020Maximum number of results to return
cursorstring—Cursor from which to return results
workspace_idrequiredstring (uuid)—The user's workspace id.