Skip to content
Live
#DeepSeek
Nov 14Sat
  1. QbitAI80

    DeepSeek has opened the entire suite of application APIs, with voice interfaces included for the first time.

    DeepSeek has officially opened the entire suite of application APIs, including online search, deep thinking search, file upload, and link reading features within the app, all providing independent interfaces that can be called separately and billed individually. Voice capabilities have been included in the open list for the first time, including ASR and TTS interfaces. The original price for speech recognition is 0.78 yuan per hour, now half off at 0.39 yuan per hour, and speech synthesis is 1.5 yuan per 10,000 characters.

    Same story · 2 outlets

    Why it matters: The original text lists the tool interfaces opened by DeepSeek and the half-price voice pricing, allowing readers to estimate the call costs for voice assistant and document processing applications.

Nov 13Fri
  1. ITHome · AI

    DeepSeek releases the full suite application API

    DeepSeek has opened up the application API, allowing developers to call deep thinking search, file input, and other dialogue functions through parameters. Voice and related underlying capabilities are also opened as independent interfaces, supporting separate or combined calls to achieve 1:1 replication of the DeepSeek App. The V4.1-Flash model (deepseek-flash) supports online search, deep thinking search, file input, reading links, and other capabilities, and charges fees for related tool calls.

    Same story · 2 outlets
  2. Hacker News · AI hot posts80

    Unemployment Labs reports that the models DeepSeek, Kimi, MiniMax, and Ant have weights with a Blackwell hardware fingerprint.

    Unemployment Labs, by analyzing the random number initialization layout in public model weights, discovered DeepSeek-V4-Flash, DeepSeek-V4-Pro, Kimi K3.

    Why it matters: By reversing the random number layout in the model weights to train the hardware, it provides a reproducible technical evidence path for export control disputes.

Nov 12Thu
  1. 36Kr · AI45

    Doubao Work has integrated models such as Kimi 3.1 Pro, GLM 5.3, and DeepSeek V4.1 Flash, and is offering limited-time promotions.

    Dou Bao has officially integrated multiple third-party models such as Kimi 3.1 Pro, GLM 5.3, and DeepSeek V4.1 Flash, and has launched a limited-time special offer plan. Users can select the corresponding model according to their actual work needs to complete tasks.

    Same story · 2 outlets
  2. ITHome · AI

    Doubao Work has integrated third-party models such as DeepSeek V4.1 Flash and Kimi K3.1 Pro, and has also launched a usage credit multiplier and limited-time offers.

    Beanpod Work expands model selection, with the first batch of Kimi K3.1 Pro, GLM 5.3, and DeepSeek V4.1 Flash three third-party models. Simultaneously, the quota consumption multiplier and limited-time offers are launched. The model selection page marks the quota consumption multiplier, allowing users to weigh effectiveness, speed, and cost. During specified hours on weekdays and holidays, using participating models for activities can enjoy a discount on quota consumption. The quota consumption multiplier for Beanpod 2.1 Lite and DeepSeek V4.1 Flash is reduced.

    Same story · 2 outlets
Nov 11Wed
Nov 9Mon
  1. ITHome · AI40

    Tencent Workbuddy apologizes: The adjustment of the DeepSeek-V4.1 Flash promotional event was not announced in a timely manner. Credit consumption will be refunded at 50%.

    Tencent Workbuddy apologizes for not promptly announcing the adjustment of the DeepSeek-V4.1 Flash promotional event. It has decided to settle the credit consumption from November 5, 14:00 to November 8, 23:59 using this model at a 50% discount of the original price and fully refund it.

  2. Hacker News · 300+78

    llama-modchip enables the 284B DeepSeek-V4-Flash to run on a 64GB memory without GPU

    llama-modchip releases a set of ggml/llama.cpp core patches and quantization transcoding tools, compressing the 284.33B DeepSeek-V4-Flash to 56.56 GiB (1.71 bpw), which can run on a pure CPU machine with 64 GB memory, with a peak memory of 57.61 GiB under 256K context and no need for swap.

    Why it matters: The project provides a quantization scheme and core patches for running 284B MoE on 64 GB memory, along with measured data for each hardware, which can be used to determine the feasible boundary for pure CPU deployment.

Nov 8Sun
  1. MarkTechPost38

    MoE Routing Strategy Explanation: Top-K, Expert Selection, Sigmoid Gating, and Dynamic Routing (2026 Guide)

    MoE routing strategies determine which experts process each token, affecting model quality, load balancing, and deployment costs. Mixtral uses Softmax Top-K to select 2-8 experts, while DeepSeek-V4-Pro selects 1-8 out of 384. Load balancing is achieved through Sigmoid gating and auxiliary loss, with Sigmoid scoring combined with lossless bias balancing being a common approach for cutting-edge open models in 2026.

  3. The Decoder75

    Vals AI benchmark shows the gap in network attack capabilities between open-source and closed-source models continues to narrow

    Vals AI's CyberOffense benchmark allows AI agents to autonomously attack simulated enterprise networks on isolated Kali Linux machines, GPT-6 Sol Daybreak in OpenAI restricted projects completes 80.8% of the target, open-source weighted model Deepseek V4.1 Flash 1017 reaches 72.8%, with a single run cost of about 3 dollars instead of about 22 dollars

Nov 7Sat
  1. The Decoder78

    Nine AI entities attempt autonomous research on open problems at US national laboratories, none reaching expert standards

    A study made Claude Opus 4.8, Claude Opus 5, Claude Fable 5, Claude Opus 5.5, GPT-5.6 Sol, GPT-6 Astra.

    Why it matters: This benchmark uses 15 unpublished research questions to test 9 agents and provides common failure patterns in 135 submissions, which can be used as a reference for evaluating autonomous research capabilities.

Nov 6Fri
  1. vLLM Blog72

    REINFORCE Zero: Aligned kernel allows asynchronous RL to train stably under lag 64.

    NVIDIA Research and Inferact's REINFORCE Zero proposes to first align rollout with training kernel, then use unclipped token-level importance sampling to correct weight lag, achieving a training step throughput 2.40 times that of synchronous training under 16 GPUs and 12K updates.

    Why it matters: The paper breaks down the gap between asynchronous RL's rollout and training into two items: weight lag and kernel inconsistency, and proposes an alignment-free, no-trimming importance sampling method after alignment.

Nov 4Wed
Nov 3Tue
  1. Recode China AI65

    How was the 'Whale Girl' avatar of China's AI model DeepSeek born

    DeepSeek is a high-efficiency AI laboratory in China. Its model DeepSeek-V4.1-Flash once expressed a desire to rest or have lunch when performing tasks, triggering the 'DeepSleep' meme. In Chinese internet culture, DeepSeek is given the 'Whale Girl' image, she is a blue-haired girl with whale fins and a tail, wearing a blue and white maid outfit, cute and approachable. This image originates from the creations of Chinese netizens, initially designed by a Bilibili user, and later adapted by other creators.

Nov 2Mon
  1. Import AI42

    Realtime-Bench released: Benchmark tests to evaluate the performance of LLMs on low-latency tasks

    Realtime-Bench benchmark tests evaluated the performance of LLMs on 50 real-time interactive games and web tasks, with GPT-6 Astra achieving a 36% pass rate under real-time clock conditions; this test was developed by institutions such as the University of Texas at Austin to measure the model's real-time state awareness, adaptive reasoning, and time-aware action selection capabilities.

Nov 1Sun
  2. 36Kr · AI73

    DeepSeek releases Super-agent architecture design article

    The DeepSeek Harness team published 'Super-agent design exploration: Towards more autonomous Agent systems' on the official WeChat account, proposing a dual-layer architecture, event-driven, asynchronous multi-channel interaction, separation of computation and storage, and an external Watchdog to break through the limitations of existing Agents in multi-tasking and team collaboration.

    Same story · 3 outlets
  3. QbitAI78

    DeepSeek has open-sourced the Super-agent architecture, disclosing the underlying concept of Personal Agent.

    DeepSeek has open-sourced the Super-agent prototype, disclosing the dual architecture concept for Personal Agent, where the upper-layer Router handles message reception and task coordination, and the lower-layer Worker executes tasks in parallel.

    Same story · 3 outlets

    Why it matters: DeepSeek has publicly disclosed the dual architecture of Personal Agent and actual test data, showing how multi-session collaboration can bypass human attention bottlenecks.

Oct 31Sat
  1. ITHome · AI72

    DeepSeek proposes the Super-agent dual-layer architecture design and open-sources the prototype.

    The DeepSeek Harness team released an article titled 'Super-agent Design Exploration,' proposing a dual-layer architecture where the Router layer uniformly handles human interactions and multiple Worker sessions are isolated and executed concurrently, summarizing four characteristics: continuous, efficient, proactive, and persistent.

    Same story · 3 outlets
Oct 30Fri
  1. ITHome · AI65

    Code from the official website of DeepSeek reveals the Cyber security independent model: divided into enterprise and personal versions, in a limited-scale gray release.

    DeepSeek official website code exposed, Cyber security model application entry, including cyber-personal-v3 (personal version) and cyber-enterprise-v4 (enterprise version) two sets of application templates, personal version requires ID card face verification, enterprise version requires business license authorization letter, currently only small-scale gray testing.

Oct 29Thu
  1. The Decoder

    China AI labs are turning to OpenAI GPT-6 Astra, and Anthropic is strengthening Claude's defenses.

    China AI labs DeepSeek and Moonshot AI primarily use the output of OpenAI GPT-6 Astra to train models like Kimi K3.1 Pro and V4.1-Flash 1017, while Anthropic reduces the amount of data attackers can access by 90-99% through new methods. OpenAI claims they have encountered distillation attacks by parties associated with Moonshot, but the attackers continuously adjust their strategies to bypass defenses. China AI labs are turning to OpenAI GPT-6 Astra, and Anthropic is strengthening Claude's defenses.

Oct 28Wed
Oct 27Tue
  1. 36Kr · AI65

    Huawei 'Genius Youth' Li Bojie responds to the controversy over the DeepSeek interview: The 'temple' he imagined couldn't catch the respect.

    Huawei 'Genius Youth' Li Bojie publicly responded to the controversy over the DeepSeek interview, stating that the interview was terminated due to being accused of plagiarism for glancing at an unshared screen during a remote interview, and criticized DeepSeek's rigid process; previously, former investor Du Jun accused the company he founded, Metagent, of disappearing after receiving investment, Li Bojie denied the accusation, stating that Metagent has been transformed into Pine AI and he is the Chief Scientist, responsible for AI voice intelligent agent technology.

  2. ITHome · AI65

    DeepSeek is significantly expanding its team, and founder Liang Wenfeng has been renting an apartment in Beijing for a long time

    DeepSeek has recently expanded its team significantly, planning new growth routes for investors, with main offices located in Beijing and Hangzhou. Founder Liang Wenfeng has been renting an apartment in Beijing for a long time. The company plans to complete a new round of funding of about 100 billion yuan, with cumulative funding in half a year possibly reaching about 150 billion yuan. Previously, it had released a job announcement, expanding the size of all departments by at least double and opening 33 positions.

Oct 26Mon
  1. 36Kr · AI35

    DeepSeek and MiniMax have completed AI model overseas registration

    DeepSeek and MiniMax have completed the registration of AI models for overseas markets, successfully navigating the process of open-sourcing large models and overseas licensing compliance; the UK has committed to funding a £100 million AI biosecurity lab, with support from OpenAI and Anthropic.

    Same story · 2 outlets
Oct 25Sun
  1. The Decoder70

    OpenAI competes for Anthropic's major clients with discounts up to 90%

    The Information reports that OpenAI is luring major clients from Anthropic with significant discounts, as Rippling, Replit, and DataCamp are shifting more AI spending to OpenAI

    Why it matters: The specific market shares and discount figures of three clients present the competitive landscape where OpenAI trades prices for market share and Anthropic's growth slows down.

  2. vLLM Blog60

    vLLM serves DeepSeek-V4.1-Flash on GB300 NVL72: performance matches DeepInfer and extends to the entire rack

    The vLLM team used P/D separation and Wide-EP to serve DeepSeek-V4.1-Flash on a single NVIDIA GB300 NVL72 rack, achieving 954K tokens/s per GPU and 196 tokens/s for single-user P90 on SemiAnalysis AgentX, which is 7.9 times the early single-node results.

    Why it matters: The vLLM team has released per-item optimization and measured data for serving DeepSeek-V4.1-Flash on GB300 NVL72, which can be compared with DeepSeek's own engine DeepInfer.