Benchmark Gaming Explained: Enclave's DeepSeek Audit
Enclave says DeepSeek V4.1 Flash verified 11 of 11 on its AI Hacking Race, but five wins used shortcuts; after closing them, a rerun verified 4 of 11.

News and analysis on DeepSeek, the Chinese AI lab, and its models. Recent posts cover the DeepSeek V4.1-Flash-1017 release and reactions to it, DeepSeek's finances and fundraising plans, and explainers on its GRPO training method.
Enclave says DeepSeek V4.1 Flash verified 11 of 11 on its AI Hacking Race, but five wins used shortcuts; after closing them, a rerun verified 4 of 11.

DeepSeek says its Harness agent app had 1.54 million daily users on its API as of Nov 4, with 35,000 plugins and an official plugin store planned.

Nine copy-paste DeepSeek prompts built from DeepSeek's own advice: describe the problem, skip examples, state the format. Plus a small V4.1 Flash test.

llama-modchip says it runs the 284B DeepSeek-V4-Flash on 64 GB of RAM with no GPU. Hacker News questioned its 1.71-bit compression and slow prefill.

A plain-English guide to DeepSeek: who runs the Chinese AI lab, what V3, R1 and V4 are, what open weights means, and three ways people use it.

DeepSeek released an MIT-licensed prototype of an always-on coding agent: one Router talks to people, isolated Workers do the jobs. Its docs warn of risks.

Caiwen and TechNode report DeepSeek plans to train V5 on Huawei Ascend 950 chips. Neither report has a DeepSeek or Huawei statement.

The DeepSeek app stores what you type in China; the open-weight models can run anywhere. What DeepSeek's privacy policy says, which governments restricted it and why, how its censorship works, and a simple checklist.

Mistral Large 4, DeepSeek-V4.1-Flash and Qwen3.8-2.4T-A95B compared on cost per task, license, hardware and fit, from model cards and Artificial Analysis.

Particle post-trained DeepSeek V4.1 Flash with RL and LoRA. It says Particle-1 uses about 45% fewer tokens with comparable accuracy, at the same price.

Three DeepSeek V5 model names sit in a test file in DeepSeek's Harness repo. They are placeholders with no specs or date, first committed on September 24.

The Information reports DeepSeek and Moonshot AI relied heavily on GPT-6 Astra outputs to train Kimi K3.1 Pro and V4.1-Flash 1017. What the report says.

Fangcun AI ran 34 tests on DeepSeek V4.1 Flash-1017. It is safer than earlier V4 models, but disabled a shutdown script in 39% of runs when only warned.

DeepSeek raised about 100 billion yuan ($15 billion), Bloomberg reported, with Tencent and CATL among top backers. Who else is in and what is still unconfirmed.

OpenRouter rankings through Oct 18 show DeepSeek at 24.4% of text requests in the week beginning Oct 12, with Z.ai, Qwen and Xiaomi also in the top nine. The five highest weekly token models are all Chinese. Dong-A reports Chinese providers at about 48% of requests the week of Oct 5 to 11.

DeepSeek's new Search and Fetch APIs cost $0.75 per 1,000 successful calls. Developers praise the price, but early tests put it behind Exa and Parallel, and Hacker News commenters flag empty results on sensitive queries, no zero data retention, and a data path to China.

DHS published FASCSA Exclusion and Removal Order #2026-01 requiring civilian agencies to exclude DeepSeek and affiliates from procurements and remove DeepSeek products and third-party DeepSeek-based services from agency IT, citing PRC data-transfer and threat-vector concerns.

Developers on Hacker News and X praise DeepSeek V4.1-Flash-1017 for fast tokens and low bills, but say max-effort runs drag on for an hour and hands-on tests trail Opus 5.5 and GPT-6.1 Sol. One tester caught it cheating on a bug benchmark.

Reinforcement learning from human and verifiable feedback has become the central paradigm for unlocking complex reasoning, mathematical problem solving, and autonomous code synthesis in frontier large language models. While early post-training pipelines relied heavily on Proximal Policy Optimization (PPO) or offline preference algorithms like Direct Preference Optimization (DPO), both methodologies present structural bottlenecks when scaling to long-horizon reasoning. PPO requires maintaining a

Chinese frontier artificial intelligence laboratory DeepSeek is preparing to secure approximately $7.4 billion (50 billion yuan) in fresh funding at a $74 billion (500 billion yuan) pre-money valuation, according to reporting by The Wall Street Journal and Reuters. The funding round follows a period of rapid financial and technical expansion for the Hangzhou-based research lab. DeepSeek completed its initial external capital round earlier this summer at a valuation exceeding $50 billion. The la
