sgl-project/sglang · error · ValueError

thinking.display is not allowed when thinking.type is…

Error message

thinking.display is not allowed when thinking.type is 'disabled'

What it means

Raised by the Anthropic-compatible endpoint's validator when thinking.type='disabled' but a thinking.display field is present. Display configuration only applies to enabled thinking, so mixing it with 'disabled' is rejected as a 400.

Solutions

1. Remove the display field when thinking.type is 'disabled'
2. Set display to null/None
3. Keep thinking only as {"type": "disabled"} when disabling

Example fix

// before
{"thinking": {"type": "disabled", "display": true}}
// after
{"thinking": {"type": "disabled"}}

if thinking["type"] == "disabled":
    thinking = {"type": "disabled"}

When it happens

Trigger: POST /v1/messages with {"thinking": {"type": "disabled", "display": true}} (or any non-null display value).

Common situations: Reusing a full thinking config dict and only flipping type to 'disabled'; SDK defaults that populate display; migrating configs between API versions where display was newly added.

Related errors

• thinking.budget_tokens is not allowed when thinking.type is…
• thinking.budget_tokens is not allowed when thinking.type is…
• thinking.budget_tokens must be >= 1024
• max_tokens must be positive
• Model is required

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/15100372c0b1df82. Report an issue: GitHub.

Appendix: source

Thrown at python/sglang/srt/entrypoints/anthropic/protocol.py:302

        if self.type == "enabled":
            if self.budget_tokens is None:
                raise ValueError(
                    "thinking.budget_tokens is required when "
                    "thinking.type is 'enabled'"
                )
            if self.budget_tokens < 1024:
                raise ValueError(
                    "thinking.budget_tokens must be >= 1024 "
                    "(got {})".format(self.budget_tokens)
                )
        elif self.type == "disabled":
            if self.budget_tokens is not None:
                raise ValueError(
                    "thinking.budget_tokens is not allowed when "
                    "thinking.type is 'disabled'"
                )
            if self.display is not None:
                raise ValueError(
                    "thinking.display is not allowed when "
                    "thinking.type is 'disabled'"
                )
        elif self.type == "adaptive":
            if self.budget_tokens is not None:
                raise ValueError(
                    "thinking.budget_tokens is not allowed when "
                    "thinking.type is 'adaptive'"
                )
        return self


class AnthropicTaskBudget(BaseModel):
    """Claude 4.7 ``output_config.task_budget`` — soft hint, not a hard cap.

    Mirrors ``BetaTokenTaskBudgetParam`` in the Anthropic SDK: ``total``
    and ``type`` are required; ``remaining`` is the client-tracked
    countdown used for compaction. The hard cap on generation is still

View on GitHub (pinned to 0132848349)