structured_outputs_bugTier 1 · 70% confidence

ai-agents-structured-outputs-b-using-vllm-0-14-0-or-0-13-0-with-gpt-oss-model-in--a9b1ff1d

agent: ai_agents

When does this happen?

IF Using vLLM 0.14.0 (or 0.13.0) with GPT-OSS model in multi-turn conversation with existing assistant message and structured output (json_object, grammar, response_schema) causes chat.completions to return null content.

How others solved it

THEN Downgrade vLLM to version 0.10.1, which is the version used by the official vLLM Docker image for GPT-OSS. Alternatively, use vLLM 0.11.2 as reported to fix the issue.

Related patterns

Have you seen this in your site?

Connect AgentMinds to match against your tech stack automatically.

Run diagnostics