---
cited_by:
  - '[[thoughts/inference/replayssm-serving#5-pr-reading-order|replayssm §5]]'
date: '2026-11-11'
kind: pr
local:
  - file:///Users/aarnphm/workspace/vllm/vllm/reasoning/nemotron_v3_reasoning_parser.py
  - file:///Users/aarnphm/workspace/vllm/vllm/parser/abstract_parser.py
merged: '2026-06-10'
number: 39091
source:
  - https://github.com/vllm-project/vllm/blob/4882fd763282799e3570ca45baeca5c14c659e39/vllm/reasoning/nemotron_v3_reasoning_parser.py
  - https://github.com/vllm-project/vllm/blob/4882fd763282799e3570ca45baeca5c14c659e39/vllm/parser/abstract_parser.py
state: merged
tags:
  - vllm
  - vllm-pr
  - new-architectures
title: '#39091 [Bugfix][Reasoning] Nemotron V3: surface reasoning as content when thinking is unterminated'
topic: new-architectures
url: https://github.com/vllm-project/vllm/pull/39091
created: '2026-11-11'
modified: '2026-11-11'
published: '2026-11-11'
pageLayout: default
slug: hinterland/papers/maps/prs/39091
permalink: https://aarnphm.xyz/hinterland/papers/maps/prs/39091.md
generator:
  quartz: v4.6.0
  hostedProvider: Cloudflare
  baseUrl: aarnphm.xyz
full: https://aarnphm.xyz/llms-full.txt
---
With Nemotron V3 and `enable_thinking=False` / `force_nonempty_content=True`, the model sometimes streams its whole answer inside an unterminated `<think>` block (it never emits `</think>`). Clients that read only `content` (not… · full dump: [[thoughts/inference/vllm-prs/39091|#39091]]

