{"version":1,"type":"story","url":"https://digestai.news/story/delta-inference-engine-runs-deepseek-v4-1-flash-natively-on-windows","json":"https://digestai.news/story/delta-inference-engine-runs-deepseek-v4-1-flash-natively-on-windows.json","markdown":"https://digestai.news/story/delta-inference-engine-runs-deepseek-v4-1-flash-natively-on-windows.md","slug":"delta-inference-engine-runs-deepseek-v4-1-flash-natively-on-windows","headline":"Delta Inference Engine runs DeepSeek-V4.1-Flash natively on Windows","summary":"The engine is a Windows port of Xolotl and Strata that uses consumer hardware, serving an OpenAI-compatible API at http://127.0.0.1:8080/v1 with a context window of up to 1 million tokens.\n\nAccording to benchmarks conducted on a system featuring an NVIDIA RTX 3090 GPU and 192 GB of RAM, the setup achieved output speeds of approximately 41 tokens per second. The system requires an NVIDIA card with at least 12 GB of VRAM, 128 GB or more of system RAM, and roughly 385 GB of free space on a high-speed PCIe NVMe SSD.\n\nThe project supports multiple quantization variants, including default ds4 Q2 files alongside formats from Unsloth and ISTA-DASLab. Setup is handled through automated scripts that check system requirements, install Python dependencies, and download checked model files directly from Hugging Face.","keyPoints":["Achieves 41 tokens per second output on an NVIDIA RTX 3090 GPU with 192 GB RAM.","Requires 12 GB+ VRAM, 128 GB+ RAM, and 385 GB free NVMe SSD space."],"whyItMatters":null,"category":{"slug":"agents","name":"Agents & Tools","url":"https://digestai.news/category/agents"},"entities":{"companies":["NVIDIA","DeepSeek","OpenAI","Anthropic","Hugging Face","Microsoft"],"models":["DeepSeek-V4.1-Flash-1017","DeepSeek-V4.1-Flash"],"people":[]},"firstPublishedAt":"2026-10-28T00:39:45Z","updatedAt":"2026-10-28T00:39:45Z","sourceCount":1,"hasPrimarySource":true,"sources":[{"outlet":"github.com","title":"DeepSeek-V4.1-Flash-1017 at 41 tok/s on one 3090","url":"https://github.com/musicurgy/delta-inference-engine","publishedAt":"2026-10-28T00:39:45Z","type":"primary","primary":true,"lead":true}],"sourceNotes":null,"discussions":[{"site":"Reddit","url":"https://www.reddit.com/r/LocalLLaMA/comments/1xh9yf7/deepseekv41flash1017_at_41_toks_on_one_3090/","points":null}],"thread":{"title":"AI Model Showdown Coding to Voice","url":"https://digestai.news/thread/recursive-releases-flint-40b-coding-agent-model-on-hugging-face","storyCount":3},"cite":{"text":"Digest AI, \"Delta Inference Engine runs DeepSeek-V4.1-Flash natively on Windows\", 28 October 2026, https://digestai.news/story/delta-inference-engine-runs-deepseek-v4-1-flash-natively-on-windows","publisher":"Digest AI","title":"Delta Inference Engine runs DeepSeek-V4.1-Flash natively on Windows","datePublished":"2026-10-28T00:39:45Z","url":"https://digestai.news/story/delta-inference-engine-runs-deepseek-v4-1-flash-natively-on-windows"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}