BerriAI/litellm · error · ValueError

Parameter {k} is not supported for model {model}. Supported

Error message

Parameter {k} is not supported for model {model}. Supported parameters are {supported_params}. Set drop_params=True to drop unsupported parameters.

What it means

LiteLLM maps OpenAI-style image-edit parameters onto Stability AI's edit API, which only supports n, size, response_format, and mask (size is translated to aspect_ratio). When map_openai_params encounters an OpenAI parameter outside that set (e.g. quality, style, user) and drop_params is False, it raises this ValueError before any HTTP request is made. The message lists the exact supported set and points to drop_params=True as the escape hatch.

Source

Thrown at litellm/llms/stability/image_edit/transformations.py:93

        # Create a copy to not mutate original - convert TypedDict to regular dict
        mapped_params: Final[dict[str, Any]] = dict(image_edit_optional_params)

        for k, v in image_edit_optional_params.items():
            if k in param_mapping:
                # Map param if mapping exists and value is valid
                if k == "size" and v in OPENAI_SIZE_TO_STABILITY_ASPECT_RATIO:
                    mapped_params[param_mapping[k]] = OPENAI_SIZE_TO_STABILITY_ASPECT_RATIO[v]
                # Don't copy "size" itself to final dict
            elif k == "n":
                # Store for logic but do not add to outgoing params
                mapped_params["_n"] = v
            elif k == "response_format":
                # Only b64 supported at Stability; store for postprocessing
                mapped_params["_response_format"] = v
            elif k not in supported_params:
                if not drop_params:
                    raise ValueError(
                        f"Parameter {k} is not supported for model {model}. "
                        f"Supported parameters are {supported_params}. "
                        f"Set drop_params=True to drop unsupported parameters."
                    )
                # Otherwise, param will simply be dropped
            else:
                # param is supported and not mapped, keep as-is
                continue

        # Remove OpenAI params that have been mapped unless they're in stability
        for mapped in ["size", "n", "response_format"]:
            mapped_params.pop(mapped, None)

        return mapped_params

    def _get_model_endpoint(self, model: str) -> str:
        """
        Get the API endpoint for a given model.

View on GitHub (pinned to 77b7c6c40c)

Solutions

Remove the unsupported parameter(s) named in the message from the image_edit call — keep only n, size, response_format, mask.
Set drop_params=True to have LiteLLM silently drop them: litellm.image_edit(..., drop_params=True) or litellm.client.LiteLLM(drop_params=True).
Set it globally with litellm.drop_params = True (or DROP_PARAMS in proxy config) when routing the same call across multiple image providers.
Replace size values like '256x256' with a size present in OPENAI_SIZE_TO_STABILITY_ASPECT_RATIO (e.g. '1024x1024', '1536x1024') so it maps to aspect_ratio.

Example fix

# before
resp = litellm.image_edit(
    model="stability/stability-image-edit",
    prompt="remove the background",
    image=open("in.png", "rb"),
    quality="hd",          # not supported by Stability
)

# after
resp = litellm.image_edit(
    model="stability/stability-image-edit",
    prompt="remove the background",
    image=open("in.png", "rb"),
    size="1024x1024",      # maps to aspect_ratio
    drop_params=True,      # or simply omit quality
)

Defensive patterns

Strategy: validation

Validate before calling

import litellm
from litellm.types.llms.stability import OPENAI_SIZE_TO_STABILITY_ASPECT_RATIO

SUPPORTED = {"n", "size", "response_format", "mask"}

def validate_stability_edit_params(params: dict) -> list[str]:
    """Return list of params that would raise; empty list means safe to call."""
    bad = []
    for k, v in params.items():
        if k not in SUPPORTED:
            bad.append(k)
        elif k == "size" and v not in OPENAI_SIZE_TO_STABILITY_ASPECT_RATIO:
            bad.append(f"size={v} (no aspect_ratio mapping)")
    return bad

bad = validate_stability_edit_params({"quality": "hd", "size": "1024x1024"})
assert not bad, f"strip these before calling: {bad}"

Try / catch

try:
    litellm.image_edit(model="stability/...", prompt=p, image=fp, **params)
except ValueError as e:
    if "is not supported for model" in str(e):
        # client-side param rejection: strip params or set drop_params and retry once
        litellm.image_edit(model="stability/...", prompt=p, image=fp, drop_params=True, **params)
    else:
        raise

Prevention

Build image-edit payloads from an allowlist (n, size, response_format, mask) instead of forwarding full OpenAI kwargs.
Set drop_params=True in multi-provider gateways so provider-specific extras never hard-fail.
Keep a unit test asserting your request dict passes get_supported_openai_params for every provider you route to.

When it happens

Trigger: Calling litellm.image_edit(model="stability/stable-image-edit", ...) with OpenAI-only kwargs such as quality="hd", style="natural", user="abc", or background="transparent". Also passing size values that are not in OPENAI_SIZE_TO_STABILITY_ASPECT_RATIO leaves 'size' unmapped and it can then fall through as unsupported. Reproduced only when drop_params is False (the default unless litellm.drop_params=True is set globally).

Common situations: Porting working OpenAI DALL-E image-edit code to a stability/ model without stripping OpenAI-specific options; a shared wrapper that injects user/quality for all providers; upgrading litellm versions where the supported-param list changed; forgetting that drop_params can be set globally via litellm.modify_params or the client constructor.

Related errors

AI-assisted analysis of BerriAI/litellm@77b7c6c40c (2026-08-18). Data as JSON: /api/errors/93a7f8d9619612e2. Report an issue: GitHub.