←Search results

Search “DeepSeek”

By relevance · Total 1346 results Clear search

Tuesday

  5. Ma Dongxi NLP

    DeepSeek Open-Sources Large-Scale Inference Framework DeepInfer Under Apache-2.0 License

    DeepSeek has open-sourced its high-performance large-scale inference framework DeepInfer under the Apache-2.0 license. The framework supports optimized kernels, model and sampler runtime, parallelism strategies such as Attention DP and balanced CP, MoE EP plus EPLB, as well as DSpark and confidence-scheduled verification, SWA bounded replay, asynchronous scheduling with a zero-overhead batch scheduler, batch-invariant inference, deterministic sampling, and structured output. Ma Dongxi NLP reposted the news and said "Respect."

    DeepSeek has open-sourced its high-performance large-scale inference framework DeepInfer under the Apache-2.0 license. The framework supports optimized kernels, model and sampler runtime, parallelism strategies such as Attention DP and balanced CP, MoE EP plus EPLB, as well as DSpark and confidence-scheduled verification, SWA bounded replay, asynchronous scheduling with a zero-overhead batch scheduler, batch-invariant inference, deterministic sampling, and structured output. Ma Dongxi NLP repost

    Matched fieldsTitle、Summary、Body

    #hot-signal#English text is being prepared.#DeepSeek#Product updates
  6. Shao MengSelected

    DeepSeek Open-Sources High-Performance Inference Framework DeepInfer

    DeepSeek has open-sourced DeepInfer, the high-performance inference framework supporting its online services, under the Apache 2.0 license. This release focuses on offline inference, is built primarily on vLLM, and draws on SGLang and TensorRT-LLM. It integrates previously separately open-sourced kernels such as FlashMLA, DeepGEMM, DeepEP, and EPLB into a complete engine, filling in system layers such as scheduling, parallelism, and speculative decoding, while supporting NVIDIA SM100 GPUs and Huawei Ascend 950 NPUs. The original article also breaks down designs including DSpark speculative decoding, YoCo SWA bounded replay, asynchronous scheduling, batch-invariant deterministic inference, and regex/JSON schema/EBNF structured output.

    DeepSeek has open-sourced DeepInfer, the high-performance inference framework supporting its online services, under the Apache 2.0 license. This release focuses on offline inference, is built primarily on vLLM, and draws on SGLang and TensorRT-LLM. It integrates previously separately open-sourced kernels such as FlashMLA, DeepGEMM, DeepEP, and EPLB into a complete engine, filling in system layers such as scheduling, parallelism, and speculative decoding, while supporting NVIDIA SM100 GPUs and Hu

    Matched fieldsTitle、Summary、Body

    #hot-signal#English text is being prepared.#Reasoning#English text is being prepared.
  7. Tianyi CuiSelected

    DeepSeek open-sources the large-scale inference framework DeepInfer under the Apache-2.0 license

    DeepSeek has open-sourced the high-performance large-scale inference framework DeepInfer under the Apache-2.0 license. The framework supports optimized kernels, model and sampler runtimes; on parallelism it provides Attention DP, balanced CP, and MoE EP + EPLB, and includes DSpark and confidence scheduling validation, SWA bounded replay, asynchronous scheduling of a zero-overhead batch scheduler, batch-invariant inference, deterministic sampling, and structured output.

    Matched fieldsTitle、Summary、Body

    #hot-signal#English text is being prepared.#Reasoning#English text is being prepared.
  8. UnslothSelected

    Unsloth Enables Local Running of DeepSeek V4.1-Flash-1017

    Unsloth announced that DeepSeek V4.1-Flash-1017 now supports local running; its 1-bit quantized version retains about 88.2% accuracy, with size reduced from 508GB to 250GB, a 52% decrease. This version requires a device with more than 128GB of memory, and Unsloth also provided a usage guide and GGUF files. DeepSeek previously stated that V4.1-Flash-1017 improved over the previous version in benchmark performance, task completion speed, and output length.

    Matched fieldsTitle、Summary、Body

    #company#English text is being prepared.#Model releases#English text is being prepared.
  9. Rohan PaulSelected

    The U.S. Department of Homeland Security requires civilian agencies to remove DeepSeek and its derivative products

    The U.S. Department of Homeland Security, pursuant to the 2018 Federal Acquisition Supply Chain Security Act, requires U.S. civilian agencies to remove DeepSeek and DeepSeek-related products, as well as third-party products and services developed using DeepSeek, from information technology systems. DHS states that DeepSeek's hosted models store data on servers in China, with every prompt and response passing through that service, and that the data may be combined with public information to form actionable intelligence targeting U.S. personnel and government programs; for model weights running on U.S. servers, the ban relies on warnings that the models may contain malware or other vulnerabilities. Exceptions require an application submitted 60 days in advance with sufficient justification, and systems in national security systems and secure isolated information systems are outside the scope of the order; the only Chinese model named in the order is DeepSeek, while Kimi and Qwen face similar distillation allegations.

    The U.S. Department of Homeland Security, pursuant to the 2018 Federal Acquisition Supply Chain Security Act, requires U.S. civilian agencies to remove DeepSeek and DeepSeek-related products, as well as third-party products and services developed using DeepSeek, from information technology systems. DHS states that DeepSeek's hosted models store data on servers in China, with every prompt and response passing through that service, and that the data may be combined with public information to form

    Matched fieldsTitle、Summary、Body

    #ai-news#English text is being prepared.#DeepSeek
  14. ArenaSelected

    DeepSeek-V4.1-Flash-20261017 Tops the Agent Arena Open-Model Leaderboard

    Arena.ai says DeepSeek-V4.1-Flash-20261017 (Max) has become the top-ranked open-source model in Agent Arena, ranking 11th overall, with a net improvement of 5.1% across more than 7.5K real sessions. The model's median cost per task is $0.20, roughly 37% of GPT-6.1 Sol's improvement margin and 27% of its cost; it ranks first among open-source models on all four signals—Confirmed Success, Praise vs. Complaint, Bash Reliability, and Steerability—and shows no tool-hallucination issues.

    Matched fieldsTitle、Summary、Body

    #editorial#Evaluations / benchmarks#Agents#DeepSeek
  17. BBC Chinese

    US Department of Homeland Security Orders Ban on DeepSeek in Government Systems

    The US Department of Homeland Security has prohibited civilian government agencies from using DeepSeek and technology derived from its models. The order was announced on October 19 but had already taken effect at the end of July. The ban applies to civilian agencies such as the Treasury Department, the State Department, and the Department of Justice, while agencies under the Department of Defense are not covered. BBC has contacted the Department of Homeland Security, DeepSeek, and the Chinese Embassy in the United States to request comment.

    The US Department of Homeland Security has prohibited civilian government agencies from using DeepSeek and technology derived from its models. The order was announced on October 19 but had already taken effect at the end of July. The ban applies to civilian agencies such as the Treasury Department, the State Department, and the Department of Justice, while agencies under the Department of Defense are not covered. BBC has contacted the Department of Homeland Security, DeepSeek, and the Chinese Em

    Matched fieldsTitle、Summary、Body

    #chinese#English text is being prepared.#DeepSeek

Monday