
  <rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
    <channel>
      <title>Chaos and Order</title>
      <link>https://www.youngju.dev/blog</link>
      <description>천천히 올바르게. AI Researcher &amp; DevOps Engineer Youngju&#39;s blog. GPU/CUDA, LLM, MLOps, Kubernetes AI workloads, and data engineering — plus mindset essays on confidence, routines, health, and sport psychology.</description>
      <language>ko</language>
      <managingEditor>fjvbn2003@gmail.com (Youngju Kim)</managingEditor>
      <webMaster>fjvbn2003@gmail.com (Youngju Kim)</webMaster>
      <lastBuildDate>Wed, 12 Aug 2026 00:00:00 GMT</lastBuildDate>
      <atom:link href="https://www.youngju.dev/tags/max-tokens/feed.xml" rel="self" type="application/rss+xml"/>
      
  <item>
    <guid>https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-context-window-vs-max-length.en</guid>
    <title>Inside vLLM (6) — Context Window vs max_model_len vs max_tokens, Fully Explained</title>
    <link>https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-context-window-vs-max-length.en</link>
    <description>If the difference between context window and max_tokens has been confusing, this single post sorts it out. Covers what each of the following actually limits, with a comparison table and real error messages: the context window of the model (a structural limit), vLLM max_model_len (an engine setting), the per-request generation caps max_tokens and max_completion_tokens, and the batch limits max_num_batched_tokens and max_num_seqs. Also covers the fact that input and output share the same budget, and how KV cache memory ends up setting the real limit. Part 6 of the Inside vLLM series.</description>
    <pubDate>Wed, 12 Aug 2026 00:00:00 GMT</pubDate>
    <author>fjvbn2003@gmail.com (Youngju Kim)</author>
    <category>vllm</category><category>context-window</category><category>max-model-len</category><category>max-tokens</category><category>llm</category>
  </item>

  <item>
    <guid>https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-context-window-vs-max-length.ja</guid>
    <title>vLLM 内部構造 (6) — コンテキストウィンドウ、max_model_len、max_tokens 完全整理</title>
    <link>https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-context-window-vs-max-length.ja</link>
    <description>コンテキストウィンドウとmax_tokensの違いが紛らわしいなら、この記事一つで整理できる。モデルのコンテキストウィンドウ(構造的な上限)、vLLMのmax_model_len(エンジン設定)、リクエストごとの生成上限であるmax_tokensとmax_completion_tokens、そしてバッチの上限であるmax_num_batched_tokensとmax_num_seqsがそれぞれ何を制限するのかを、比較表と実際のエラーメッセージで整理した。入力と出力が同じ上限を分け合うという事実、KVキャッシュのメモリが実質的な上限を決める構造まで扱う。vLLM内部構造シリーズ第6回。</description>
    <pubDate>Wed, 12 Aug 2026 00:00:00 GMT</pubDate>
    <author>fjvbn2003@gmail.com (Youngju Kim)</author>
    <category>vllm</category><category>context-window</category><category>max-model-len</category><category>max-tokens</category><category>llm</category>
  </item>

  <item>
    <guid>https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-context-window-vs-max-length</guid>
    <title>vLLM 내부 구조 (6) — 컨텍스트 윈도우, max_model_len, max_tokens 완전 정리</title>
    <link>https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-context-window-vs-max-length</link>
    <description>컨텍스트 윈도우와 max_tokens 차이가 헷갈린다면 이 글 하나로 정리됩니다. 모델의 컨텍스트 윈도우(구조적 한계), vLLM의 max_model_len(엔진 설정), 요청당 생성 상한인 max_tokens와 max_completion_tokens, 그리고 배치 한도인 max_num_batched_tokens와 max_num_seqs가 각각 무엇을 제한하는지 비교표와 실제 에러 메시지로 정리했습니다. 입력과 출력이 같은 한도를 나눠 쓴다는 사실, KV 캐시 메모리가 실질 한계를 정하는 구조까지 다룹니다. vLLM 내부 구조 시리즈 6편.</description>
    <pubDate>Wed, 12 Aug 2026 00:00:00 GMT</pubDate>
    <author>fjvbn2003@gmail.com (Youngju Kim)</author>
    <category>vllm</category><category>context-window</category><category>max-model-len</category><category>max-tokens</category><category>llm</category>
  </item>

    </channel>
  </rss>
