{"version":1,"type":"story","url":"https://digestai.news/story/deepseek-v4-1-flash-reduces-ai-agent-memory-usage","json":"https://digestai.news/story/deepseek-v4-1-flash-reduces-ai-agent-memory-usage.json","markdown":"https://digestai.news/story/deepseek-v4-1-flash-reduces-ai-agent-memory-usage.md","slug":"deepseek-v4-1-flash-reduces-ai-agent-memory-usage","headline":"DeepSeek releases V4.1-Flash with 1M context and MIT license","summary":"DeepSeek AI has released DeepSeek-V4.1-Flash, a multimodal Mixture-of-Experts model designed to address the memory bottlenecks of long-horizon agent workloads. The model features a 552B parameter backbone with 196B additional Engram parameters and supports a 1M-token context window. By activating only 8B parameters during prefill and 16B during decode, it significantly reduces computational overhead compared to previous iterations.\n\nThe release introduces several architectural innovations, including a Causal Encoder-Decoder structure that halves prefill compute and Compressed Sparse Attention 2 (CSA2), which reuses KV cache data across layers. These optimizations reduce the global KV cache footprint to 890 bytes per token, roughly 437 times smaller than DeepSeek-V1. The model also utilizes FP4 quantization for the main KV cache, further cutting storage requirements.\n\nDeepSeek-V4.1-Flash is available under an MIT license with open weights on Hugging Face, supporting deployment via vLLM, SGLang, and Transformers. The team reports that the model outperforms competitors like Opus-5 and GPT-5.6 Sol on specific agentic benchmarks, such as Terminal-Bench 2.1 and DeepSWE v1.1, while maintaining efficient resource usage for enterprise and research applications.","keyPoints":["DeepSeek-V4.1-Flash features 552B parameters and a 1M-token context window with MIT-licensed open weights.","New CSA2 architecture and FP4 quantization reduce KV cache footprint to 890 bytes per token.","Model outperforms Opus-5 and GPT-5.6 Sol on Terminal-Bench 2.1 and DeepSWE v1.1 agentic tasks."],"whyItMatters":"This release lowers the hardware barrier for deploying million-token context models, enabling cost-effective long-horizon agents. The MIT license and open weights accelerate adoption in enterprise and research settings, challenging proprietary closed-source alternatives.","category":{"slug":"models","name":"Generative AI & Models","url":"https://digestai.news/category/models"},"entities":{"companies":["DeepSeek AI"],"models":["DeepSeek-V4.1-Flash","DeepSeek-V4-Flash","DeepSeek-V1","Opus-5","GPT-5.6 Sol"],"people":[]},"firstPublishedAt":"2026-09-10T12:40:51Z","updatedAt":"2026-09-14T12:00:09Z","sourceCount":6,"hasPrimarySource":true,"sources":[{"outlet":"MarkTechPost","title":"DeepSeek AI Released DeepSeek-V4.1-Flash with 1M Context, FP4 KV Cache, and Cross-Layer Attention Reuse","url":"https://marktechpost.com/2026/09/10/deepseek-ai-released-deepseek-v4-1-flash-with-1m-context-fp4-kv-cache-and-cross-layer-attention-reuse","publishedAt":"2026-09-10T07:31:01Z","type":"press","primary":false,"lead":true},{"outlet":"KDnuggets","title":"Why DeepSeek-V4.1-Flash Is Such an Exciting Open Model Release","url":"https://kdnuggets.com/why-deepseek-v4-1-flash-is-such-an-exciting-open-model-release","publishedAt":"2026-09-14T12:00:09Z","type":"press","primary":false,"lead":false},{"outlet":"github.com","title":"DeepSeek V4.1 Flash at 41 tok/s on 8× A40 on TensorSharp","url":"https://github.com/zhongkaifu/TensorSharp","publishedAt":"2026-09-12T08:47:09Z","type":"primary","primary":true,"lead":false},{"outlet":"Unite.AI","title":"Baseten Adds DeepSeek-V4.1-Flash to Model APIs With 1M-Token Context","url":"https://unite.ai/baseten-adds-deepseek-v4-1-flash-to-model-apis-with-1m-token-context","publishedAt":"2026-09-11T22:24:02Z","type":"press","primary":false,"lead":false},{"outlet":"finance.yahoo.com","title":"SK Hynix, Samsung Shares Dip In Korea After DeepSeek Debuts AI Tech That Uses Less Memory – Micron, SanDisk Hold Up","url":"https://finance.yahoo.com/technology/ai/articles/sk-hynix-samsung-shares-dip-071811226.html","publishedAt":"2026-09-11T05:30:00Z","type":"press","primary":false,"lead":false},{"outlet":"The Decoder","title":"New Deepseek model V4.1-Flash cuts memory needs for AI agents","url":"https://the-decoder.com/new-deepseek-model-v4-1-flash-cuts-memory-needs-for-ai-agents","publishedAt":"2026-09-10T12:40:51Z","type":"press","primary":false,"lead":false}],"sourceNotes":null,"discussions":[{"site":"Reddit","url":"https://www.reddit.com/r/LocalLLaMA/comments/1we7ay2/deepseek_v41_flash_at_41_toks_on_8_a40_on/","points":null}],"thread":{"title":"DeepSeek Launches Open Long Context Agent Model","url":"https://digestai.news/thread/deepseek-v4-1-flash-reduces-ai-agent-memory-usage","storyCount":3},"cite":{"text":"Digest AI, \"DeepSeek releases V4.1-Flash with 1M context and MIT license\", 10 September 2026, https://digestai.news/story/deepseek-v4-1-flash-reduces-ai-agent-memory-usage","publisher":"Digest AI","title":"DeepSeek releases V4.1-Flash with 1M context and MIT license","datePublished":"2026-09-10T12:40:51Z","url":"https://digestai.news/story/deepseek-v4-1-flash-reduces-ai-agent-memory-usage"},"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"}