https://www.youngju.dev/blog 천천히 올바르게. AI Researcher & DevOps Engineer Youngju's blog. GPU/CUDA, LLM, MLOps, Kubernetes AI workloads, and data engineering — plus mindset essays on confidence, routines, health, and sport psychology. ko fjvbn2003@gmail.com (Youngju Kim) fjvbn2003@gmail.com (Youngju Kim) Wed, 12 Aug 2026 00:00:00 GMT https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-context-window-vs-max-length.en https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-context-window-vs-max-length.en 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. Wed, 12 Aug 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) vllmcontext-windowmax-model-lenmax-tokensllm https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-context-window-vs-max-length.ja https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-context-window-vs-max-length.ja 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。 Wed, 12 Aug 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) vllmcontext-windowmax-model-lenmax-tokensllm https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-context-window-vs-max-length https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-context-window-vs-max-length 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. Wed, 12 Aug 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) vllmcontext-windowmax-model-lenmax-tokensllm