{"type":"post","id":"how-to-make-flash-models-even-faster-optimizing-131ad4","slug":"how-to-make-flash-models-even-faster-optimizing-131ad4","title":"How to make flash models even faster: optimizing DeepSeek V4.1 Flash 1017","description":"Baseten says it optimized inference for DeepSeek V4.1 Flash 1017 by iterating on more than 80 kernels and applying kernel fusion, Programmatic Dependent Launch, and host-loop changes.","retrieval_nugget":"Baseten says it optimized inference for DeepSeek V4.1 Flash 1017 by iterating on more than 80 kernels and applying kernel fusion, Programmatic Dependent Launch, and host-loop changes. It reports 991 output tokens per second on Artificial Analysis. The work illustrates how serving efficiency can depend on model-specific engineering as well as model choice.","published_at":"2026-11-11","updated_at":"2026-11-11","record_date":"2026-11-11","date_kind":"published_at","topics":["models"],"entities":["Baseten","DeepSeek V4.1 Flash 1017","SGLang","DeepInfer","Artificial Analysis","NVIDIA","GLM-5.3","Cline"],"related_patterns":["agent-economics-moves-to-completed-work"],"novelty_label":"new_fact","what_changed":"Baseten describes a new optimization of DeepSeek V4.1 Flash 1017, reporting 991 output tokens per second and the lowest time per task on Artificial Analysis.","prior_refs":["https://newruntime.com/posts/change-log-deepseek-api-docs-a5a148/","https://newruntime.com/posts/vllm-blog-serving-deepseek-v4-1-flash-at-073f14/"],"source_url":"https://www.baseten.co/blog/how-to-make-flash-models-even-faster-optimizing-deepseek-v41-flash-1017-fast","source_urls":["https://www.baseten.co/blog/how-to-make-flash-models-even-faster-optimizing-deepseek-v41-flash-1017-fast"],"schema_version":"newruntime-agent-readable-v0.2","stable_id":"post:how-to-make-flash-models-even-faster-optimizing-131ad4","status":"published","visuals":[],"editor":{"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]},"routes":{"html":"https://newruntime.com/posts/how-to-make-flash-models-even-faster-optimizing-131ad4/","markdown":"https://newruntime.com/posts/how-to-make-flash-models-even-faster-optimizing-131ad4.md","json":"https://newruntime.com/posts/how-to-make-flash-models-even-faster-optimizing-131ad4.json"},"markdown_url":"https://newruntime.com/posts/how-to-make-flash-models-even-faster-optimizing-131ad4.md","next_reads":[{"type":"topic","path":"/topics/models/","url":"https://newruntime.com/topics/models/","title":"Models","reason":"Explore the models topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/models.md","json_url":"https://newruntime.com/topics/models.json"},{"type":"collection","path":"/posts/","url":"https://newruntime.com/posts/","title":"Field Notes","reason":"Open the latest editorial analysis.","media_type":"text/html","json_url":"https://newruntime.com/posts.json"},{"type":"collection","path":"/sources/","url":"https://newruntime.com/sources/","title":"Sources","reason":"See the original sources this archive cites.","media_type":"text/html"}]}